How Institutional Framework Shapes Bank Efficiency in Sub-Saharan Africa

How Institutional Framework Shapes Bank Efficiency in Sub-Saharan Africa P. Louis Arnaud TAMINI 1 University of Strasbourg & LaRGE Research Center August 2017 Abstract In this paper, we investigate whether the institutional framework influences banking efficiency in Sub-Saharan Africa, especially in UEMOA and CEMAC countries. Because of the heterogeneity across institutional framework, banks may operate in different environments, conditioning their ability to supply credit to the economy. We found that the institutional framework determine bank efficiency. Indeed, banks in UEMOA are more efficient than those of CEMAC when the model doesn t include the environmental conditions. However, after controlling with the specific environment in each country, the gap between the two zones becomes non-significant. Moreover, additional investigates show that banks in UEMOA are operating in a homogeneous environment, while in CEMAC the environment is heterogeneous. Overall, the strength of legal right index, the index for enforcement of contracts, and the regulatory quality explain the differences in loans production within the two areas. Reforms that promote the private sector, and protect the rights of borrowers and lenders are essential in order to improve bank technical ability to produce credit. JEL Classification: G21; G28; G32; F36 Keywords: Africa, Bank Efficiency, Credit, Institutional Framework 1 ptamini@unistra.fr

1. Introduction Despite the undertaken reforms in the past decades, the financial sector of many Sub-Saharan Africa (SSA) countries still remains underdeveloped compared to other developing regions. For instance, examining the financial development in Africa in international comparison, Beck and Cull (2014) find that the median private credit to Gdp ratio is 34% in non-african developing countries, but only 18% in Africa. The weakness of the financial inclusion is also one of the characteristics of SSA financial sector. Only 21% of the firms have a line of credit or loan from a formal financial institution, while this indicator reaches 43% in non-african developing countries (Demirgüç-Kunt and Klapper, 2012). Moreover, the financial sector is widely bank-based. In many countries, the Non-bank financial institutions (NBFIs) are marginal, almost non-existent. In CEMAC 1 countries, the banking system accounts for more than 80% of the financial assets. In this context, where the financial sector is widely dominated by banks, it is important to investigate their efficiency to ensure that they fulfill their role especially since financial development contributes to growth (Levine, 2005). Moreover, in the literature, institutions have always been identified as fundamental tools for nations stability, growth and progress (Smith, 1759; Veblen, 1899; North, 1991). Indeed, the institutional framework helps understanding the gap between countries in terms of economic performance. Hall and Jones (1999) show that the gap in output per worker between countries can be explained by differences in institutions and governance. Acemoglu et al. (2001), by studying the impact of institutions on income per capita, confirm that institutions matters. In fact, once they control for the effect of institutions, countries in Africa and those closer to the equator do not have lower incomes. Our paper is related to this extensive literature. More specifically, we focus on how institutional framework determines bank efficiency in Sub- Saharan Africa countries. With the growth of the banking sector, the institutional reforms, and the founding of subregional organizations, SSA countries have aroused interest for researchers regarding banking efficiency (Kirkpatrick et al. 2008; Hauner and Peiris, 2008; Mlambo and Ncube, 2011). We apply on the UEMOA 2 and CEMAC banks, the methodology developed by Dietsch and Lozano-Vivas (2000). Basically, Dietsch and Lozano (2000), by comparing bank efficiency between France and Spain, suggest that the definition of a common frontier has to incorporate country-specific environmental conditions. Indeed, banks within these two countries evolve in different environments. Thus, in some extent a cross-country comparison would be biased if 2

the environmental conditions are not accounted for. We proceed here to a cross-regional comparison. In fact, the institutional and environmental conditions that shape the financial sector in SSA might also be determined by the integration policies pursued at the regional level. Considering SSA, the region is notably structured along two sub-regional economic and monetary unions: CEMAC and UEMOA. Thus, in addition to consider the impact of institutional and environmental conditions on the technical efficiency of SSA banks in a comparative perspective, we analyze each of the two organizations separately. Indeed, our methodology provides an original way to characterize the integration of the banking industry in each of these zones. The results show that on average banks are more efficient in UEMOA than CEMAC when we do not account for the environmental conditions. However, when the model includes the specific institutional framework of each country, there are no differences between the two zones in terms of efficiency. Additional investigates confirm that countries within the UEMOA zone are homogeneous, while those of CEMAC are heterogeneous. Overall, the quality of institutions but also the macroeconomic conditions determine banking efficiency in SSA. The more the strength of legal right index, the index for enforcement of contracts, and income are high, the more is the increase in technical efficiency. Conversely, the more the regulatory quality and the financial development are high, the less is the increase in technical efficiency. In this context, reforms that promote the private sector and guarantee the rights of lenders and borrowers are necessary in order to increase banks technical ability to produce credit. In our knowledge, this paper is the first which brings to light the important influence of the institutional framework in determining banking efficiency in SSA countries. Moreover, our paper also provides a new approach to evaluate the level of financial integration in UEMOA and CEMAC. The remainder of the paper is organized as follows. Section 2 provides the background of the research question. Section 3 focuses on the methodology and data. In section 4, the results are presented, and the section 5 concludes. 3

2. Background 2.1. Literature review Studies about banking efficiency in SSA are few. Unlike for developed countries, the first investigations in the field have been conducted in the 2000s. Before, the weakness of financial depth and specifically the lack of data were hindrances to such investigations. Much of them were first oriented on the nexus between competition and banking efficiency in SSA (Hauner and Peiris, 2008; Buchs and Mathisen, 2005; Mlambo and Ncube, 2011). Indeed, in the 80s and 90s most of SSA countries have implemented policies in order to restructure their financial sector. The aim was to promote the financial development and therefore boosting growth and reduce poverty. These policies have been accompanied by the entry of foreign banks in many SSA countries, and therefore an increase of competition in the banking sector. During this period, the focus was on the relationship between competition and efficiency. Apart from papers related to the nexus competition - banking efficiency, studies about bank efficiency in SSA countries can be grouped in three classes: (1) studies which assess bank efficiency without any consideration of the environment influence, (2) studies which first determine the level of efficiency and then in a second stage investigate the determinants of these efficiency levels by using some bank specific factors and external environmental variables as explanatory variables, and (3) studies which account for the potential impact of environment in building the efficient frontier. Among studies which assess banking efficiency without taking into account the potential impact of the environment, we can quote Ncube (2009) and Kamau (2011). The first one by measuring the efficiency of the four larger banks and the four smaller banks in South-Africa find that bank size is negatively correlated with cost efficiency. Kamau (2011) takes a look on intermediation efficiency and productivity in the Kenyan post liberalization banking sector. In terms of ownership, the results show that foreign banks are the most efficient, then the private domestic banks, and lastly the public local banks. Regarding the size, large banks appear to be more efficient than medium and small ones. Moreover, there is an improvement in terms of productivity suggesting that Kenyan banks have integrated the technological change from liberalization reforms. The question of the efficiency s determinants has also been widely addressed. Generally, this question is handled in two steps: first, bank efficiency levels are determined, and in a 4

second stage, variables related to banks, macroeconomic conditions, legal and institutional framework, are used to explain the efficiency levels. With regard to bank specific factors, the literature is unanimous concerning their impact on bank efficiency. Kirkpatrick et al., (2008) 3 find that bad loans and high capital ratios contribute to both cost efficiency and profit efficiency. Concerning size, large and medium banks appear to be the most cost-efficient in SSA banking industry (Kirkpatrick et al., 2008, Kiyota, 2011), while the smallest banks are more profit efficient (Kiyota, 2011). Overall, there is evidence that foreign banks are the most efficient (Kirkpatrick et al., 2008; and Kiyota, 2011). However Kablan (2009a), by examining the efficiency with regard to ownership, finds that private local banks are the most efficient, followed by foreign ones, and lastly public banks. The results from Kiyota (2011) also suggest that Pan-African banks are more profit efficient that non-sub-saharan Africa foreign ones. Besides bank specific factors, the environmental variables also determine bank efficiency in SSA countries. The first set of those external environmental variables are the macroeconomic conditions. Indeed, variables such as income, inflation, and financial depth are emphasized by the literature as significant in determining efficiency scores (Chen, 2009 and Kiyota, 2011). Thus, a stable macroeconomic framework contributes to higher banking efficiency. Moreover, the market structure also influences bank efficiency (Chen, 2009 and Kablan, 2009a). Then, variables related to the legal framework and the quality of institutions are the other external factors that determine efficiency levels. Chen (2009) found that stronger legal institutions and enforcement of contracts, as well as political stability and government effectiveness are beneficial to banking efficiency. In the mentioned cross-country studies, none of them consider the environmental conditions in generating the common efficient frontier. By resorting to the one-step procedure conceived by Battese and Coeli (1995), Kablan (2009b) addresses this gap. With this method, the impact of variables that condition cost-efficiency is integrated to the cost frontier. Therefore, the obtained levels of efficiency are supposed to account for the potential influence of the environmental conditions. However, it is important to point out that only two environmental variables were included: the level of income, and the percentage of rural population. In this paper, we extent this approach by integrating in the efficient frontier other variables related to the legal framework and the quality of institutions. 5

2.2. The UEMOA and the CEMAC: between similarities and heterogeneity This article investigates the impact of environmental variables on bank efficiency in SSA. Moreover, in terms of economic and financial approach, two main zones arise in Africa: the UEMOA and the CEMAC. Indeed, these two zones are the most successful attempt of economic and monetary integration in the continent. In each zone, common policies are applied within the economic integration. Therefore, two characteristics common to both zones can be point out: on the one hand, considering the countries within each zone, they are theoretically homogeneous due to the implementation of common policies; on the other hand, considering both zones, they have similarities regarding their banking systems, but also differences related to the macroeconomic and legal framework. Given these characteristics, the UEMOA and the CEMAC provide us the framework for examining how bank efficiency can be influenced by the environment. Established in 1994, the West African Economic and Monetary Union (UEMOA) is an economic and monetary organization which is made up of 8 member States: Benin, Burkina Faso, Ivory Coast, Guinea Bissau, Mali, Niger, Senegal, and Togo. These countries have a common Central Bank, the BCEAO 4, and a single currency, the Franc CFA. Among its objectives, the BCEAO is committed to develop and apply the common monetary policy for all the state members. It also ensures the stability of the banking and financial system of the community. Because of some historical reasons, the BCEAO has an agreement with the Banque de France, which allows fixing the Franc CFA exchange rate with euro. In return, the BCEAO has to deposit more than 50% of its foreign exchange reserves at the Banque de France. The Central African Economic and Monetary Community (CEMAC) is exactly the equivalent of the UEMOA, but it gathers the central African countries. It has been founded in 1999, five years after the UEMOA. It has 6 member States, namely: Cameroun, The Central African Republic, The Republic of Congo, Gabon, Equatorial Guinea and Chad. Like the UEMOA, the CEMAC countries also have a common Central Bank, the BEAC 5, and a single currency, the Franc CFA 6, also linked to the Banque de France, and to the euro. Thus, the UEMOA and the CEMAC have some similarities: they have the same currency linked to the Banque de France and, as economic and monetary organizations, they are pursuing the same objectives. In addition to follow very similar monetary policies, the 6

financial sectors within the CEMAC and the UEMOA show similar structures. Financial markets are underdeveloped in both zones. Moreover, their financial sectors are widely dominated by banks. The Non-bank financial institutions (NBFIs) are almost non-existent. For instance, in the CEMAC the banking system accounts for more than 80% of the financial assets. In both zones, a large part of the banking system is held by foreign investors: at least 50% of the assets (BAD 7, 2010; Allen, 2011; IMF, 2016). One of the common characteristics to both zones is the concentration of the banking system. In each of the CEMAC's countries, about 70% of the assets are held by the three largest banks. In the whole UEMOA zone, five banks account for 50% of the banking assets (IMF, 2013; IMF, 2016). The access to formal banking services is a hindrance in both zones. For example, on average, in the CEMAC countries, lees than 15% of adults are bank accounts holders (Beck and Cull, 2014). Overall, despite the reforms in financial sectors during the past decades, the financial development is still weak and, the access to formal financial services is limited. Moreover, the ownership structures and the governance of SSA banks are similar. Indeed, most of the existing banks in the two zones are subsidiaries of French groups, and the executives are from French schooling (Kablan, 2009b). However, the size of the financial sector is smaller in CEMAC zone, and its depth is higher in UEMAO zone. In 2013, for CEMAC 50 banks were registered in the 6 member states. At the same period, in UEMOA the banking system was made up of 114 banks. The UEMOA countries have a level of financial development higher than CEMAC ones. Indeed, the ratio of credit to Gdp is 10% in CEMAC countries, while in the UEMOA this ratio reaches 20%, either the double (IMF, 2013; IMF, 2016). Nevertheless, the banking industries in the two zones appear to be very close, which allows assuming that the banking technology is the same in all the countries considered here, thus allowing the computation of a common production frontier and efficiency scores. However, major structural differences, related to the demography, the economy, and the business environment, can be outlined. In terms of demography, the differences between the UEMOA and the CEMAC countries are substantial. First of all, the total population is obviously higher in UEMOA: around 106 million inhabitants versus 46 million in CEMAC. The gap between the two areas reaches on average 60 million of inhabitants. Concerning the density of population, the same situation is observed. Indeed, the density of population in UEMOA is more than the triple of that of the CEMAC, respectively 63 inhabitants / km² against 19 inhabitants / km² (World Bank, 7

2013).This could produce some significant differences in the demand for banking products and services among households. Significant economic differences also exist between the UEMOA countries and those of the CEMAC. First, the economic specialization is different between the two zones. In fact, among the 6 countries of CEMAC, 5 are oil producers. Only the Central African Republic is an exception to the rule. Oil represents 41% of the GPD of the region and 86% of the goods exports (FMI, 2012). On the other side, the UEMOA countries are predominantly exporters of agricultural products, namely cotton, coffee and cocoa. As consequences, the UEMOA countries are poorer than those of the CEMAC. Second, concerning inflation, it is higher in CEMAC. Over 2009-2012 period, the average annual inflation was 2.1% in UEMOA versus 3.3% in CEMAC (FMI, 2013). Moreover, in UEMOA, since 16 March 2012, the reserve requirement ratio is 5% for the whole banks. In CEMAC there is no common reserve requirement ratio for all the member States. In fact, because of banking liquidity differences between member States, reserve requirement are applied on case-by-case from one country to the next (Banque de France, 2008). With regard to the business environment, compared to CEMAC, the UEMOA offered the best environment for doing business (Doing Business, 2013). The ease of doing business is evaluated by the Distance to Frontier (DTF). According to the definition of the World Bank, "the distance to frontier score aids in assessing the absolute level of regulatory performance and how it improves over time. This measure shows the distance of each economy to the frontier, which represents the best performance observed on each of the indicators across all economies in the Doing Business sample since 2005." This indicator is ranked from 0 to 100, where 0 represent the lowest performance, and 100 the frontier. In 2013, on average, the DTF was 44.57 for the UEMOA countries, and 40.96 for CEMAC ones. To summarize, on the one hand, banks in UEMOA and CEMAC have the same technology. On the other hand, these banks operate in different environments that are likely to be shaped at the sub-regional level. In this context, to properly measure and compare their level of efficiency, we have to control for environmental factors, in order to take into account the conditions in which banks evolve. 8

3. Methodology and data 3.1. Methodology Sub-Saharan Africa has one of the most underdeveloped banking systems in the world, although large disparities exist between countries (Honohan and Beck, 2007; Beck and Cull, 2014). For instance, countries like Kenya or Nigeria have appeared to be major financial centers with a banking system closer to that of emerging countries. By contrast, the countries of the UEMOA and CEMAC have less developed banking systems. These two groups of countries can be bracketed together in terms of financial development. Within these economic and monetary unions, common economic and financial policies are implemented to facilitate convergence. Thus, banking technology is the same in UEMOA and CEMAC, respectively. However, banks operate in different environments. As stated previously, these differences might be related to the demography, the macroeconomic conditions, the legal framework, the business environment, and so on. In this context, any comparison between the two zones will be biased, if we don t control for the specific environment of each country. We follow the same methodology than Dietsch and Lozano-Vivas (2000) to assess the influence of environment on banking efficiency in UEMOA and CEMAC areas. We adopt the parametric approach 8 (Aigner et al. 1977; Meeusen and van den Broeck, 1977) to conduct our analysis. In the context of banking efficiency, parametric approaches consist generally in estimating an efficient frontier and then in measuring differences between the point at which each bank is operating (X-Efficiency) and the efficient frontier. Structural approaches have the advantages to discern between random errors and inefficiency even if they make some assumptions about their distribution. But in return, they impose a particular functional form for the frontier. In the literature, for these kinds of analysis, the Cobb-Douglas and the Logarithmic Transcendental (Translog) production functions are usually used. The Translog function is a generalization of the Cobb-Douglas function. In this paper, we opt for the Translog production function because it offers a flexible (second-order) functional form. We resorted to the intermediation approach to select inputs and output. Indeed, one of the characteristics of African banks is the low production of credit despite the predominance of commercial banks. Banks in the area are unable to ensure the financing of private sector. For instance, in Africa, only 22% of firms have access to credit, and 45% of the firms consider the access to funding as an obstacle to their development (Demirgüç-Kunt and Klapper, 2012). Moreover, only 74% of deposits are converted to credits, versus 109% for the others 9

developing countries. In this context, it is important to investigate the loan production in SSA and, more specifically banks technical capacity to transform the collected deposits into loans. Therefore, throughout the regressions, we consider a translog stochastic production function (Christensen et al, 1973):,,, 6!" # = $%+$&' ( )!* #( $+$ 1 2 $&$& / (0)!* #( )!* #0 $$ +$&3 4 5 #4 $+$7 # $$$$$$$$$$$$$$$$$$$$819 (-. (-. 0-. 4-. Where Y i is the Production of the i-th bank; X ik (k = 1, 2, 3) the input k of the i-th bank; 5 #4 (p=1 to 5) the environmental variables p of the i-th bank. Note that, when the estimates are 6 done without account for the environment, we drop the term$$: 4-. 3 4 5 #4 in (1). 7 # = ; # <$> # represents the error term of the i-th bank; V i are traditional random variables and are assumed to be iid. N(0, σ² v ); U i 0, are random variables that are supposed to account for the technical inefficiencies in the production process. In our model, U i are independent and identically distributed exponential with scale parameter σ u. The density function for U i is given by: $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$?8@9 =$ 1 A B expc< @ A B D$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$829$$$$$$$$ V i and U i are distributed independently of each other, and of regressors. Thus, their joint density function can be written as the product of their individual density: 1 $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$?8@E F9 = $ $JKL${< @ < F² G2HA B A I A B 2AM }$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$8N9$$$$ I Model (1) represents a Translog production function with one output and three inputs. As stated previously, inputs and output are determined by the intermediation approach because in UEMOA and CEMAC, banks are specialized in collecting deposits, and granting loans. Market activities are few. Thereby, the production is approximated by Loans, while inputs are made up by Borrowed Funds to what we add Labor and Capital. These variables are defined more precisely in Table 2. Using model (1), technical efficiencies (TE i ) are determined. The Technical Efficiency is defined as the ratio of observed output to maximum feasible output given the effects of random shocks out of the control of each bank. The Technical Efficiency has necessarily 10

values between one and zero. Thus, when the bank achieves its maximum feasible, the technical efficiency is equal to 1; otherwise, TE i < 1. Mathematically, this definition is formalized by the following expression: "!" =! #($ ; %).&$'{* }!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(4) Where #($ ;%) is the production frontier; &'${* } captures the effect of random shocks on each producer; ", $ and % are already defined above. As stated before, the error term of the translog production function is made up by two components: + =, -!/. The main problem is to distinguish between!,!et!/, and more precisely extracting the information on / contained in +. As a solution, Jondrow et al. (1982) proposed to consider the expected value of / conditional on +.!They showed that if /!!are distributed exponential, the conditional distribution of / given + is : #(0 +) =! #(01+) #(+)!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!= 2 (0 - :<)? 3567 8 9(- :<!exp >- ) 57? @!!!!!!!!!!!(A) 7 8 Where :< =!-+ - ( B C D ); 9 is the standard normal cumulative distribution. BE #(01+) is the joint density function of 0!and!+, and is given by :!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!#(01+) =! &$' I- J - F (0 K 3?GH E H C H E?HD +)? L!!!!!!!!!!!!!!!!(M) C F #(+) is the marginal density function of +, and is obtained by integrating 0 out of #(01+): P! Q!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!#(+) =!N #(01+)O0!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!=! R T!9 S- BE BC U!B C BE V&$'!I W BE!U! B C D DB E D!L!!!!!!!!!!!!!!!!!!!!!(X) #(0 +) is distributed as N + (:<17 8? ), and its mean is given by the following expression : 11

(! " # " )=$%' & +*, -./012 3 45 6 7/ 1 3 (8) 2 68$$ 45 With, the standard normal density distribution function. After obtained the estimates of! ", the Technical Efficiency of each Bank is measured by : ############################################################$% " =&'({)û " }#################################################################*9+# Jondrow et al. (1982) defined û i as %*! ", " +. Thus, by substituting this expression in equation (9), we obtained the following measure of TE: ####################################$% " =exp{)%*! ", " +}###################################################*10+ 3.2. Data Given our methodology, we resort to two kinds of data: data from banks balance sheets and income statements, and data related to macroeconomic conditions and legal framework at the country level. Both cover the period 2007-2013 and consist of pooled cross-sectional data. Data on banks are provided by Bankscope. We take into account all the banks available for UEMOA and CEMAC countries. However, for Guinea-Bissau (UEMOA) and Equatorial Guinea (CEMAC), there is a lack of data. Therefore, we drop these two countries. On UEMOA our sample is made up by 73 banks, and concerning CEMAC, we have 33 banks in the sample (Table 1). These data from banks are used to determine inputs and output. Here, we consider one output (Loans) and three inputs. The first input is approximated by borrowed funds. Borrowed funds are made up by total customers deposits plus deposits and short term funding. The two remaining inputs are those which are traditionally used in the production, namely Labor and Capital. Labor is approximated by Personal Expenses, and Fixed Assets represent Capital. Table 2 presents average values of inputs and outputs, in thousands of USD, for UEMOA and CEMAC from 2007 to 2013. The average values of inputs and output suggest that banks in UEMOA are bigger than CEMAC ones. Over 2007-2013, the collected deposits and the granted loans are high in UEMOA, compared to 12

CEMAC. Both Personal Expenses and Fixed Assets are higher in UEMOA than CEMAC, almost the double. Overall, we notice that in UEMOA and CEMAC, the financial intermediation is low. Environmental variables are provided by World Bank Indicators and World Bank s Worldwide Governance Indicators (WGI). Throughout the regressions, we take into account three major variables related to the quality of institutions in building the common frontier. Indeed, in SSA countries, we consider the institutional framework as a fundamental cause in explaining the gap in loans production. These three institutional variables are: the strength of legal right index, the index for enforcement of contracts, and the regulatory quality. The Strength of Legal Right Index is included in our estimates because it plays an important role in regulating lenders and borrowers relationship. According to World Bank, the Strength of Legal Right Index measures the degree to which collateral and bankruptcy laws protect the rights of borrowers and lenders and thus facilitate lending. The index ranges from 0 to 12, with higher scores indicating that these laws are better designed to expand access to credit. Basically, the UEMOA and CEMAC countries have inherited an institutional transfer from France during colonization. Thus, their legal systems are close (Bruyas, 2008). However, in order to attract investors, these countries have implemented some changes in their laws during the last decades. These changes are related to collateral and bankruptcy laws. In this context, this variable could be important in our approach. The Index for Enforcement of Contracts is defined by the number of procedures. It is the number of independent actions, mandated by law or courts that demand interaction between the parties of a contract or between them and the judge or court officer (World Bank, 2013). Countries which have a low number of procedures have the best effectiveness in terms of enforcing a contract. Conversely in countries with high number of procedures, the court system is slow and the plaintiff has to wait a long time before to get actual payment after filing a dispute. On average, in CEMAC countries, the number of procedures for enforcement a contract is 42 (table 3). This number is quite homogeneous for all the countries within the area, with the lowest number of procedures in Gabon (38) and the highest in Central African Republic (43) and Republic of Congo (44). In UEMOA, on average, plaintiff has to go through 38 procedures before getting actual payment after filing a dispute. However, there are differences across countries: in Ivory Coast the number of procedures is 33, while it is almost 13

44 in Senegal (World Bank, 2013). Overall, the index for enforcement of contracts is low in both zones, as the number of procedures is high. Lastly, the Regulatory Quality is also considered. This variable reflects perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development (Kaufmann et al., 2010). The indicator is ranged from -2.5 (weak) to 2.5 (strong). Overall, regulatory quality is weak in both zones, as it is on average -0.55 in UEMOA and -0.91 in CEMAC. All the countries within the two areas have a non-positive rating, below zero (World Bank, 2013). These rating are consistent with 2013 Doing Business Report where UEMOA and CEMAC countries are classified in the 50 worst performing countries in the world. A high level of regulatory quality potentially allows bank to evolve through an attractive environment. Therefore, this should have a positive impact on their efficiency. However, besides these three variables related to the institutional framework, we also control for macroeconomic conditions. To do so, we include two variables which are often used in the literature. First the income variable, represented by GDP per Capita. It is one of the most used variables in empirical studies and especially those related to efficiency (Dietsch and Lozano-Vivas, 2000; Chen, 2009; Kablan, 2009b). We control for this variable because it could be a major determinant for bank loans and deposits level. Overall, the level of income is higher in CEMAC because countries are beneficing of oil revenues. However, there is relatively some heterogeneity between countries in this zone, comparing to UEMOA. Then, to account for financial development, we use a traditional indicator namely domestic credit to private sector in percentage of GDP (Credit to private sector/gdp). This variable has been used by Chen (2009) 9 on efficiency assessment in Sub-Saharan African middle-income countries. He found a positive impact with banking efficiency. A higher level of financial development could potentially improve bank performance and efficiency. On average the level of financial development is about 20.34% in UEMOA area versus 9.51% in CEMAC one (Table 3). The whole UEMOA countries have a level of financial development higher than 10%. Contrary, in the CEMAC area only Cameroon and Gabon have a level of financial development reaching 10%. Especially in Chad the financial depth is extremely low, about 4% (World Bank, 2013). 14

Regarding the environmental variables, all the indicators are weak for both zones, but they seem to be more favorable in UEMOA than CEMAC. Indeed, except for income, all the indicators are higher in UEMOA. 4. Results We first present estimates on a frontier common to all CEMAC and UEMOA countries (4.1), then on UEMOA and CEMAC separate frontiers (4.2). We conclude this section by characterizing which institutional variables mostly condition bank efficiency (4.3). 4.1. Results on a common frontier The production function is first estimated on a frontier common to all CEMAC and UEMOA countries, by supposing that the efficiency is determined by banking technology only. So, we assume that the environment in which banks operate does not matter. The results show that on average the technical efficiency is higher in the UEMOA countries (0.8151) than in the CEMAC zone (0.7343). The difference between the two zones is high (8.08%) and statistically significant at the 1% confidence level (Table 6). Thus banks in UEMOA are technically more efficient than in CEMAC. Before controlling explicitly for the environment, we test the difference between the two zones by introducing a dummy variable for the UEMOA. Thus the reference zone is the CEMAC. This estimate allows capturing zone effects. To some extent, the zone effect is here assimilated to an aggregated environmental index. The dummy variable is significant at the 1% confidence level (Table 4, Model 1), confirming that there are some institutional and economic differences between the two zones. Next we introduce the environmental variables. After controlling for the environment, bank efficiency level still remains higher in UEMOA than CEMAC. The average technical efficiency is now 0.8233 for UEMOA banks versus 0.8177 for CEMAC ones. However, the average gap of 0.56% is no longer significant (Table 6). 15

By controlling for the environmental conditions of each zone, we set a comparable basis. This allows saying that in the case of UEMOA and CEMAC countries, bank efficiency in terms of lending is sensitive to the institutional framework. More precisely, these results highlight the differences between the two zones. Concerning the UEMOA, controlling for the environmental conditions has no material effect on average bank technical efficiency, as the difference after having controlled is not significant (Table 5). This result suggests that in UEMOA countries, banks are operating in a homogeneous environment. Conversely, regarding the CEMAC zone, there are some significant differences in efficiency levels once we controlled for the specific environment of each country. Thus, in CEMAC zone, banks are operating in a heterogeneous environment. On average, in both zones, there is an increase in efficiency level after having controlled for the environmental effects (Table 5). However, this increase is not systematic for all the banks. The inclusion of environmental variables in the common frontier has a negative impact on some banks. Indeed, in UEMOA zone at least 25% of the banks become less efficient, while in CEMAC, only less than 5% of the banks are negatively impacted. Considering the environmental variables, except regulatory quality, they are all significant (Table 4, Model 2). Income appears with a negative sign in our regressions. Financial development has a positive impact on banks efficiency. This result was expected. A high level of financial depth contributes to more performance and efficiency. The result is also consistent with Chen (2009) in the case of cost efficiency. Index for enforcement of contracts has the expected sign (negative). This variable represents the number of procedure to enforce a contract. A high level of index for enforcement of contracts tends to reduce technical efficiency. Strength of legal right index, which measures the degree to which collateral and bankruptcy laws protect the rights of borrowers and lenders and thus facilitate lending, has a negative sign. Overall, efficiency is conditioned by the specific context in which banks are evolving. Therefore, to proper compare bank from two countries or groups of countries, we have to define the best common frontier by integrating the local conditions in which banks are operating. 16

4.2. Results on separate frontiers In the previous section, we investigated how the environment conditions bank efficiency on the whole sample, by estimating a frontier common to both zones. Here, we closely examine the situation in each zone by estimating two separate frontiers. This approach allows deepening the analysis within each zone. 4.2.1. Results on the UEMOA regional frontier On its regional frontier, the average technical efficiency in terms of lending for UEMOA banks without environmental variables is 0.8216 (Table 8). By taking into account the potential influence of environment in the model, we observe that the technical efficiency reaches 0.8320. The difference of 0.0104 is not statistically significant (Table 8). Thus, by controlling for the environmental conditions, we find that in the UEMOA the efficiency is constant. This finding is also consistent with the one on the common frontier. Regarding the environmental variables, except strength of legal right index, we find that none is significant (Table 7, Model 1). At this stage, this result tends to point out that in the UEMOA zone, countries have close characteristics. In order to deepen the analysis, we perform a model including country dummy variables. The estimates are done relatively to Ivory Coast. Like with the environmental variables, none of the dummy variables is significant, validating that there is no institutional differences between UEMOA countries (Table 7, Model 2). On the whole, these results on the UEMOA regional frontier confirm our previous finding on the common frontier concerning the fact that the UEMOA countries are homogeneous. The banking technology is similar through the zone, and the operating conditions seem to be the same whatever the country. Thus, countries within the UEMOA have very close characteristics in terms of banking technology as well as governance practices. In some way, our results confirm those of Sy (2007) who found that financial integration in the UEMOA area is well advanced when it comes to markets participants facing the same rules. Moreover, Diarra (2014) also finds that the UEMOA countries are convergent with regard to total outstanding debt and tax pressure. In 2013 10, among the first four convergence criteria in force, three were respected by all the eight UEMOA countries. Then, the UEMOA countries have begun their convergence and this is already noticeable in the banking sector. 17

To summarize, in UEMOA, the efficiency of banks, in terms of lending, is more related to bank management, i.e. the way banks combine their inputs to produce outputs than to institutional and environmental conditions. Moreover, the implemented policies in the context of financial integration are effective as the countries within the area are homogeneous. 4.2.2. Results on the CEMAC regional frontier The results on the CEMAC regional frontier, without environmental variables, show that on average, banks have a technical efficiency of 0.7705 (Table 10). The CEMAC banks still have scope to improve their efficiency in terms of lending. Moreover, the technical efficiencies are heterogeneous within the area. For instance, the difference between the country with the highest efficiency level (0.8537 for Central African Republic), and the country with the weakest one (0.6686 for the Republic of Congo) is 18.51% (Table 10). The standard deviation of the efficiency scores for all CEMAC banks is 13.20%. The technical efficiency reaches on average 0.8120 when the model includes the environmental conditions. The difference with the previous efficiency score is 4.15%, and is statistically significant (Table 10). Thus these results point out the heterogeneity of the CEMAC zone. We carry out some additional estimates by setting dummy variables for CEMAC countries (Table 9, Model 2). The whole countries dummy are significant confirming that bank are operating in heterogeneous environments. Within the area, countries don't have the same characteristics so that banks are operating in different environments. The financial integration is limited between countries throughout the zone. Regarding the set of environmental variables, they are all significant, except regulatory quality (Table 9, Model 1). Moreover, the signs are consistent with those observed on the common frontier. Income, index for enforcement of contracts, and strength of legal rights index tend to influence negatively lending production in CEMAC. Then a high level of financial development is beneficial for banks. Overall, the results for the CEMAC regional frontier validate those on the common frontier: in the CEMAC zone, countries are heterogeneous and the institutional framework determines bank efficiency in terms of lending. 18

4.2.3. Explaining the gap of efficiency in lending production: which variables are the most determinant? The results, both on separate and common frontiers, show that in the UEMOA and the CEMAC, the institutional variables play an important role in conditioning the technical efficiency levels. The knowledge of the particular influence of each of these variables could be useful to make recommendations for the reform of institutional framework. In this last section, we try to well understand in which way each variable affect the level of technical efficiency when we include it in the model. In other words, we characterize the marginal impact on bank technical efficiency of controlling for a given environmental variable. The following model is estimated: '!""_#$ =!" +!# () $ %& +!* (12) With!%&, the set of five Environmental Variables used through the regressions;,% -./!0!,% -./12. 3 is the technical efficiency level with (without) environmental Variables; 4566_,% is the difference in technical efficiency between the frontiers given by:!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4566_,% =!,% -./ 7!,% -./12.!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!0893! We estimate several models (Table 11). The models 1, 2 and 3 concern the full sample. The model 4 concerns only the UEMOA zone, while the last one is for the CEMAC. The results are consistent through the whole specifications. Indeed, all variables are significant with the same signs, except the model 5 where index for enforcement of contracts and regulatory quality are not significant. Strength of legal rights index, index for enforcement of contract and income have positive signs. First, when we control for income, the increase in technical efficiency is more important particularly in countries with high income levels. Thus, by not accounting for this variable, we underestimate bank efficiency in high-income countries. This explanation is in line with our previous findings. In fact, we find that the increase in efficiency score is about 8.34% for CEMAC countries, versus only 0.83% for UEMOA ones (Table 5). All things being equal, a part of this increase in CEMAC countries can be explained by the fact that on average the income is high in this zone, relatively to UEMOA. 19

Second, concerning strength of legal rights index, and index for enforcement of contract, the sign of the regression coefficient is also positive. Once we control for legal rules, the gain in technical efficiency is higher especially in countries with strong legal framework. In other words, the more the legal rules are powerful and guarantee, the more is the increase in technical efficiency. Then, in countries with low legal rules, the inclusion of those variables has a limited impact on bank efficiency. This suggests that improvement of these indicators would have a limited impact on bank technical efficiency in countries with weak legal institutions, at least in the short run. For these three variables, the positive regression coefficient reflects a larger underestimation of bank technical efficiency in loan production for countries with high values of these indicators, all things being equal. On the other side, financial development, and regulatory quality have negative signs: the more financial development and regulatory quality are high, the less is the increase in technical efficiency. Therefore, after having accounted for these two variables, the increase in technical efficiency is low particularly in countries where the financial depth is high and the regulatory quality strong. In countries where financial development and regulatory quality are better, the impact on bank efficiency by controlling with these variables is limited. Conversely, there is an important impact for countries where their levels are low. As policy implications, reforms that promote the quality of legal framework, and sustain the growth are necessary in order to improve bank technical ability to supply effectively loans to the economy. Moreover, our results overall suggest that ignoring the institutional environment in efficiency measurement leads generally to an underestimation of average technical efficiency. However, all institutional and environmental variables do not have the same impact on bank technical efficiency. 20

5. Conclusion In this paper, we investigated how the institutional framework shapes bank efficiency in Sub- Saharan Africa and more specifically in UEMOA and CEMAC countries. The study allows understanding the differences in technical efficiencies between countries, especially in terms of lending. We found that banks in the UEMOA zone are more efficient than those of CEMAC when the model doesn t include the environmental conditions within each zone and country. However, when the model takes into account the macroeconomic and institutional conditions, the gap between the two zones reduced significantly, becoming statistically non-significant. In others words, there is no difference between UEMOA and CEMAC banks in terms of technical efficiency, when the comparison includes the specific conditions in which banks are operating in each country. This result points out the fact that bank efficiency level is not only related to the combination of inputs and output, but also depends on the specific environmental conditions within countries. More specifically, in the case of UEMOA and CEMAC countries, the institutional framework plays an important role in explaining the differences in loan production. Additional estimates on separate frontiers confirm that in UEMOA zone the environment is homogeneous, while in CEMAC there are some institutional differences between countries. Overall, in UEMOA and CEMAC countries, the macroeconomic conditions, and the quality of institutions determine bank technical efficiency in terms of lending. Yet, in those countries and in Sub-Saharan Africa in general, the access to credit is a hindrance for households and small and medium enterprises. In this context, it is the responsibility of the leaders to take on the necessary reforms in order to set up a favorable institutional environment for the banks. More specifically, reforms that promote the private sector, and protect the rights of lenders and borrowers are required in order to increase banks technical ability to produce credit. 21

1 In French, CEMAC refers to «Communauté Économique et Monétaire de l Afrique Centrale». In English, CEMAC is called Central African Economic and Monetary Community. 2 In French, UEMOA refers to «Union Économique et Monétaire Ouest Africain» In English, UEMOA is called West African Economic and Monetary Union 3 As stated in the paper, the findings have to be taken with caution because the sample is dominated by two countries namely Nigeria and Kenya which represent in total 64% of the banks 4 BCEAO means in French, «Banque Centrale des Etats de l Afrique de l Ouest.» 5 BEAC : in French «Banque des Etats de l Afrique Centrale» 6 Even if they have the same name, «the Franc CFA BCEAO» and «the Franc CFA BEAC» are different and are not interchangeable : 1 euro = 655,957 Francs CFA BCEAO and 1 euro = 655,957 Francs CFA BEAC 7 Banque Africaine de Développement 8 We resort to Stochastic Frontier Analysis (SFA) 9 Chen (2009) used Deposits to GDP rather than Credits to GDP. 10 See International Monetary Fund (IMF) report (2013) concerning the UEMOA countries. 22

References Acemoglu, D., Johnson, S., and Robinson, J. A. (2001), The Colonial Origins of Comparative Development: An Empirical Investigation, American Economic Review 91, 1369-1401. Aigner, D., Knox Lovell C.A., and Schmidt, P. (1977), Formulation and estimation of stochastic frontier production functions models, Journal of Econometrics 6, 21-37. Allen, F., Otchere, I., and Senbet, L. (2011), African Financial Systems: A Review, Review of Development Finance 1, 79-113. Asongu, S. (2012), Bank Efficiency and Openness in Africa: Do income level matter? Review of Finance and Banking 4, 115-22. Bamba, L. (2004), «Analyse du Processus de Convergence dans la zone UEMOA», World Institute for Development Economic Research (UNU-WIDER), Working Papers, UNI- WIDER Research Paper RP2004/18 Banque Africaine de Développement (2010), «Intégration du Secteur Financier dans Trois Régions d Afrique : Comment l intégration financière régionale peut soutenir la croissance, le développement et la réduction de la pauvreté», Banque de France (2008), «Rapport Zone franc». Banque de France (2002), «Zone Franc», Note d information N 127. Battese, G. E., and Coelli, J. (1988), Predicting firm-level technical efficiencies with a generalized frontier production function and Panel Data, Journal of Econometrics 38, 387-399. Beck, T. and Honohan, P. (2007), Making Finance Work for Africa, World Bank, Washington, DC. Beck, T. and Cull, R. (2014), Banking in Africa, In Berger A., Molyneux P. et Wilson J.(éd), The Oxford Handbook of Banking, 2nd edition. Bruyas, J. (2008), Les Institutions de l'afrique Noire Moderne, L Harmattan, Paris. 23