DETERMINANTS OF COMMERCIAL BANKS PROFITABILITY IN SUB-SAHARAN AFRICA. Abstract

Transcription

1 DETERMINANTS OF COMMERCIAL BANKS PROFITABILITY IN SUB-SAHARAN AFRICA By: Munyambonera Ezra Francis Abstract This research investigated some of the key determinants of commercial banks profitability in Sub-Saharan Africa. The study used data from balance sheet as well as standardized financial accounts derived from the Bank-Scope International bank database. An unbalanced panel data set for a sample of 224 commercial banks from 42 countries, for the period 1999 to 2006 was utilized. A cost efficiency model was employed to generate specify the bank profitability function. The random effects estimator was utilized to estimate the model. Results confirm the importance of bank level factors such as; assets, capital adequacy, operational efficiency, and liquidity; and macroeconomic factors such as growth in GDP and inflation in explaining bank profitability in SSA. Further, the results provide evidence that that the banking sector over the study period had persistence in profitability behavior towards equilibrium. By studying the determinants of banks performance in Sub- Saharan Africa, the study provided additional knowledge about SSA commercial banking sector that is important for policy making. 1

2 DETERMINANTS OF BANK S PROFITABILITY IN SUB-SAHARAN AFRICA 2.0 Introduction During the last two decades, the banking sector in Africa and in the rest of the developing world has experienced major transformation in its operating environment. In a number of countries, financial sector reforms have been implemented. In these reforms, the role of commercial banks has remained central in financing economic activities in the various segments of the markets especially in Sub-Saharan Africa. Panayiotis et al.(2005), Naceur and Goaied (2001; 2003) among others, showed that both external as well as domestic factors have contributed to growth in performance of SSA banks in the last two decades. These studies further suggested that, given the importance of commercial banks in Africa, better understanding of the determinants of performance was important. The few studies on Africa and other developing world have shown that there are many factors that influence bank profitability. Commercial banks profitability for most of the SSA countries has been about 2 percent over the last 10 years and compares significantly with other developing world, but lower than the developed world. A major research question is why commercial banks in SSA have remained less competitive and less profitable, despite financial sector reforms of 1980s and 1990s, aimed at improving their efficiency. The motivation behind this study was that there is little information about commercial banks performance in SSA that would be important for policy guidance of the sector. 2

3 2.1 Problem Statement Despite financial sector reforms in Africa since the 1990s with an aim of improving profitability, efficiency and productivity, commercial banks performance has remained poor with substantial gaps in service delivery to private agents. There is sufficient empirical evidence that poor performance is manifest into low performance of bank indicators including: high levels of credit risk to private agents, poor quality loans, limited and or inadequate capitalization, operational inefficiencies, higher incidences of non-performing loans, higher levels of liquidity risk; among others. Although these are mentioned as constraint areas affecting SSA banks performance, they are based on a few studies and nonelaborate methods to generate sufficient conclusions. This study there became an extension of the few studies undertaken on SSA with a view of generating sufficient information on commercial banks. The study adopts the fundamental indicators that influence banks performance in general and have been utilized in most studies available. These observations are collaborated in the literature and empirical studies contained in this paper. Nissanke and Aryeetey (1998) and Aryeetey et al. (1997) demonstrated that continuous poor performance of banking systems in Africa could be partly explained by the high degree of financial market fragmentation and limited access to basic payment services or savings accounts. On the other hand, Nissanke and Aryeetey (2006); Dermerguc-Kunt and Huizingha (2001); and Bikker and Hu (2002) revealed that African banks have not been widely studied and was therefore difficult to inform policy on readily efficient banks in the continent without sufficient data. A similar view was reached by Roland (1997), Eichengreen and Gibson, (2001), Goddard et al. (2004), Gibson (2005), Bonaccorsi di Patti 3

4 and Hardy (2005), Berger et al. (2005), and Nakane and Weintraub (2005), that more data was required on African banking systems to inform policy. All these studies, among others observed that more understanding on African banking sector performance was important. World Bank (2005) also emphasized the need to undertake deeper analysis of financial sector performance in SSA, where performance has not been impressive. 2.2 Objectives of the study The main objective of the study was to investigate the determinants of commercial banks profitability in SSA for the period 1999 to The research should help to draw some implications for policy that improves performance of the sector in the sub-region. The study utilized both bank level as well as macroeconomic factors to measure profitability. 2.3 Justification for the Study Empirical evidence clearly shows that studies focusing on Sub-Saharan Africa s financial sector are still scanty and limited. Even those which have been carried out point to a need for further investigation of the factors which that have continued to cause poor financial performance in the sub-region, notwithstanding the reforms. Most of the evidence in regard to commercial banks performance largely focus on the developed economies environments and the conclusions of may not be useful for Africa financial sector planning. According to literature, the studies on commercial banks profitability would provide more elaborate and current information that is important for policy for the sector and also scholarly literature. 4

5 2.4 Hypotheses Testing The major hypothesis of this study was to evaluate weather bank level as well as macroeconomic factors are important in explaining commercial banks profitability in SSA. The report is organized as follows. Chapter one presents the background and study motivation. In Chapter II, the determinants of banks profitability are explored. In chapter three, the determinants of banks total factor productivity are investigated. The analysis was based on unbalanced panel for 224 banks, operating in 42 countries for the period 1999 to The study follows an extensive literature that focuses on bank level as well as macroeconomic factors as main determinants of bank profitability. Bank data was derived from the Bank-Scope and macroeconomic data from International Finance Statistics database. A dynamic panel specification was adopted to estimate the determinants of banks profitability in SSA. Return average assets and net interest margin were adopted as alternative measures of bank profitability and considered dependent variables; while bank-specific and macroeconomic factors; as explanatory. Bank-specific factors include: growth in bank assets, capital adequacy, asset quality (credit risk), operational efficiency, and liquidity ratio; while macroeconomic variables include inflation and growth in GDP. The main conclusion of the study is that bank-specific as well as macroeconomic factors explain the variation in SSA commercial banks profitability. The report is structured as follows. Section 2.1 presents the introduction Section 2.2 5

6 describes the literature review of some of the important studies on bank performance. In section 2.3, empirical model and methodology are discussed. In section 2.4, empirical findings are discussed, while in 2.5, conclusions and some implications for policy are indicated. 2.5 Review of Literature on Bank Profitability This section explores the empirical literature and the various methods used in studying commercial bank s profitability. In theory, determinants are categorized into three indicators: bank-specific, industry-specific and macroeconomic. Bank specific indicators include: growth in bank assets, capital adequacy, operational efficiency, and liquidity. The common measure for industry-specific representative used in the various studies is bankconcentration. While on the other hand, the key macroeconomic variables include: growth in GDP, GDP-per-capita inflation expectation, interest rate and its spread. The empirical evidence provides the various methods employed in studying bank profitability using these determinants. Much of the empirical literature agrees that bank level as well and macroeconomic factors largely influence bank profitability. There is however limited evidence that industry-specific factors have any influence on bank profitability. It is against this background that the study utilized only bank level and macroeconomic factors to estimate profitability. 6

7 2.5.1 Bank-specific determinants In trying to understand commercial banks performance in Sub-Saharan Africa, studies on profitability have largely focused on returns on bank asset or equity and net interest margin. Also, traditionally, the impact on banks performance has been measured by bank-specific factors such as capital adequacy, credit risk, liquidity risk, market power and regulatory costs. More recently, research seems to have focused on the impact of macroeconomic factors on banks performance. In all these studies, however, the literature reveals that Sub- Saharan Africa is less studied and therefore would require more information on banking sector for better planning for the sector. This study therefore, was an attempt to address the gap of information on SSA commercial banking sector. In investigating bank profitability,short (1979), Bourke (1989), Molyneux and Thornton (1992), Demirguc-Kunt and Huizinga (2000) and Goddard et al (2004), among others, applied linear models to explain performance. Linear models have however been criticized for employing inconsistent variables and generating inefficient results. All this evidence points to the fact that that more data would be required to understand banks performance in the developing countries. Cornett and Tehania (1992), Mercia, et al. (2002), Toddard, et al. (2004), Guru et al.(2001) and Panayiotis et al.(2006) show that bank profitability is a function of internal and external factors. Internal factors include bank-specific; while external factors include both industry-specific and macroeconomic factors. According to this literature, there are six standard key bank-specific indicators that are widely used to study banks: profitability; 7

8 capital risk; asset quality; operational efficiency; and growth in bank assets. Industry specific factors include: ownership, bank size, bank concentration index; while macroeconomic factors include interest rate, interest rate spread, inflation and per-capita income and growth in GDP. Most of these factors are included in this study to estimate bank profitability for SSA banks. Appling the General Method of Moments (GMM) technique to a panel of Greek banks for the period 1985 to 2001 period, Panayiotis et.al. (2006) discovered that bank profitability persist to a moderate extent. Persistence suggests that departures from perfectly competitive market structures may not be large. The study further shows that all bank-specific determinants, with the exception of size, influence bank performance in the anticipated way. The study on Malaysian banks by Guru et al. (2001) also show that efficient management is among the most important factors that explain high bank profitability. Extending a similar study to SSA, therefore, generates comparative results. Al-Haschimi (2007) investigates the determinants of bank net interest margin in 10 SSA countries, and applies an accounting decomposition model as well as panel regressions. Further found out that credit risk and operational inefficiencies explain most of the variation in net interest margins across the region, with macroeconomic factors, having less influence on performance Smirlock (1985) found a positive and significant relationship between size and bank profitability. Short (1979) discovered that size is closely related to capital adequacy of a 8

9 bank since relatively large banks tend to raise less expensive capital and, hence, appear more profitable. Using similar arguments, Bikker and Hu (2002) and Goddard et al. (2004), among others; linked bank size for small to medium size banks to capital and profitability. Molyneux and Thornton (1992), among others; found a negative and significant relationship between the level of liquidity and profitability. In contrast, Bourke (1989) reported an opposite result; while the effects of credit risk on profitability could be negative (Miller and Noulas, 1997). There is also an extensive literature based on the idea that an expense-related variable should be included in a profit function. For example, Bourke (1989) and Molyneux and Thornton (1992) found a positive relationship between better-quality management and profitability. Athanasoglou, et al. (2006) study on the South Eastern European banking industry over the period 1998 to 2002, suggested a new approach in understanding bank profitability Industry-specific determinants Another strand of literature emphasizes the importance of market structure and bank specific variables in explaining performance heterogeneities across banks. This literature is based on the structure-conduct-performance (SCP) paradigm and is also applicable to contestable markets, firm-level efficiency, and the roles of ownership and governance in explaining banks performance (Berger, 1995; Berger and Humphrey, 1997; Bikker and Haaf, 2002; Goddard et al., 2001; Goddard et al., 2004; and Molyneux et al., 1996). It was explained that, the extensive empirical evidence does not provide conclusive proof that 9

10 bank performance is influenced either by concentrated market structures and collusive price setting behavior or superior management and production techniques. Bank efficiency levels were found to vary widely across banking sectors (Altunbaş et al., 2001; Schure et al., 2004). Smirlock (1985) and Berger (1995) investigated the profit structure relationship in banking, providing tests of the aforementioned hypotheses. To some extent, the relative market power hypothesis was verified; since there was evidence that superior management and increased market share (especially in the case of small to medium sized banks) raise profits. In contrast, weak evidence was found for the efficient structure hypothesis. Explained that efficiency not only raises profits, but may lead to market share gains and, hence, increased concentration, so that the finding of a positive relationship between concentration and profits could be a spurious result due to correlations with other variables. Bourke (1989), Molyneux and Thornton (992) argued, instead that increased concentration is not the result of managerial efficiency, but rather reflects increasing deviations from competitive market structures, which leads to monopolistic profits. Consequently, concentration should be related to bank profitability. These studies further questioned whether the ownership status of a bank is related to its profitability or not. However, little evidence is found to support the theory that privatelyowned institutions will return relatively higher economic profits (Short, 1979). In contrast, Bourke (1989) and Molyneux and Thornton (1992) revealed that ownership status could be irrelevant for explaining profitability. Eichengreen and Gibson (2001) analyzed bank- 10

11 specific and market-specific profitability determinants of Greek banks for the period 1993 to 1998; using a panel not restricted to commercial banks. The results revealed that industry-specific variables such as concentration ratio and market share have positive and significant influence on bank profitability. Recent literature also emphasizes the importance of changes in macroeconomic conditions on bank profitability. All this evidence provides sufficient information that bank concentration may not be a larger factor in explaining banks profitability behaviour Macroeconomic determinants The last group of profitability determinants deals with macroeconomic control variables. The common variables include inflation rate, the long-term interest rate and rate of economic growth (Panayiotis et al., 2005). More recently, a number of studies emphasize the relationship between macroeconomic variables and bank risk. Allen and Saunders (2004) provided evidence of the importance of macroeconomic factors in determining the profitability of banks in the sampled Rovell (1979) introduced the issue of the relationship between bank profitability and inflation. Noted that the effect of inflation on bank profitability depends on whether bank wages and other operating expenses increase at a faster rate than inflation. The question is how mature an economy is so that future inflation is accurately forecasted to enable banks manage their operating costs. Perry (1992) observed that the extent to which inflation affects bank profitability depends on whether inflation expectations are fully anticipated. An inflation rate fully anticipated by the bank management implies that banks can 11

12 appropriately adjust interest rates in order to increase their revenues faster than their costs and thus acquire higher economic profits. Bourke (1989) and Molyneux and Thornton (1992) found a positive relationship between inflation and bank profitability. Saunders and Schumacher (2000) apply a model of Ho and Saunders (1981) to study the determinants of interest margin in six European Union and US banks during the period 1988 to They further established that macroeconomic volatility and regulations have a significant impact on bank interest margin. The result pointed out an important trade off between ensuring bank solvency, as defined by high capital to asset ratio, and lowering cost of financial services to consumers, as measured by lower interest rate margin. Bourke (1989), Molyneux and Thornton (1992), Demirguc-Kunt and Huizinga (2000) and Bikker and Hu (2002) identified possible cyclical movements in bank profitability. Bikker and Hu. (2002) established that bank profits are positively correlated with movements in the business cycle. Afanasieff et al. (2002), Guru et al. (2002) and Naceur and Goaied (2001; 2003), and Barajas et al. (1999) study on emerging countries (Brazil, Colombia, Malaysia and Tunisia) documented significant effects of financial liberalization on banks performance. Afanasieff et al. (2002) also made use of panel data techniques to uncover the main determinants of bank performance in Brazil and found out that macroeconomic variables such as GDP growth rate, inflation expectations are important in determining bank profitability over time. Neeley and Wheelock (1997) also explored the profitability of 12

13 sampled US commercial banks and found a positive impact of per-capita income on profitability. Overall, empirical review for this research provides back ground information of bank profitability. There is ample evidence of comprehensive account of developed countries and a few of developing ones, but less of SSA, signifying the requirement for further research on the sub-region. Empirical findings provide support that bank profitability is influenced by both internal, sector specific as well as macroeconomic factors. This study utilized some of the identified key variable indicators that were identified in earlier studies to have an influence on commercial banks profitability. Further, there is proof that in these studies, more of static models than dynamic have been applied to study banks profitability. Dynamic panel models therefore deem to be considered as important specifications for providing more rigorous and efficient results on banks performance in the context of SSA, which is an extension of these studies. This study on SSA commercial banks; becomes an important step in providing more empirical information on the sub-region s commercial banking sector; using dynamic panel models. These methods are expected to generate more efficient results. 2.6 Conceptual Framework In this section, the theoretical basis of the relationship between banks profitability and explanatory bank-specific as well as macroeconomic factors is presented. From the theoretical relationship, the conceptual model that was used to estimation is presented. 13

14 2.6.1 Theoretical basis for the model A cost efficiency model was employed to measure bank profitability. This approach was adopted from the work done by Marcia, et.al. (2002), Marcos (2003), Kang and William (2005), Goddard, et. al. (2004), and Panayiotis et al. (2005; 2006) on bank efficiency in developed and a few developing economies. These studies employed dynamic panel specifications to estimate the determinants of bank's performance. These approaches are understood to generate reliable estimates on larger samples (Evanoff and Israelvich, 1991; Wheelock and Wilson, 1999). In measuring bank's performance, the key bank indicators aggregated into as well as some macroeconomic factors, such as inputs and outputs were utilized to measure performance Cost efficiency model for generating the profitability function The cost efficiency frontier is a technical efficiency concept based on a production function that is used to measure bank cost efficiency. Cost efficiency is derived from the cost function and is a modified form of Cobb-Douglas function. This provides information on how close (or far) bank costs are from the best practice, producing the same output under similar conditions. Cost efficiency therefore reflects the position of particular bank relative to the cost frontier. A stochastic cost frontier function is presented, where C (.) is a suitable functional form; lnc i = C(y i,w i, β ) + ν i +u i ; I = 1,2..., N...(2.1) Where c i is the observed cost of production; y i is the logarithm of output quantity; w i is the 14

15 vector of logarithms of input prices, β is a vector of unknown parameters to be established; v i is the random error term and u i is the non-negative inefficiency effect. Coelli, Rao and Battese (1998) showed that inefficiency factor u i is included because the cost frontier represents minimum costs. 1 The random error v i accounts for measurement errors and other random factors. Inefficiency factor accounts for both technical inefficiency and allocative inefficiency. The random error and the inefficiency term are assumed to be multiplicatively separable from the cost frontier Berger and Mester (1997). Efficiency measurement techniques differ in how they separate the composite error term v i + u i, i.e. how they distinguish the inefficiency term from the random error. Battese and Coelli (1992) employed a stochastic frontier specification of the cost efficiency model to study Central and Eastern Europe to estimate bank cost efficiency. Likewise, Marko (2006) employed a common frontier function in analyzing efficiency gaps between East and West Europe banks following integration. In the same vein, this methodology was adopted to study SSA commercial banks in this study. Using the basic model specification (2.1), a log linear generalized production function framework was utilized for estimating bank profitability. The structural model is presented in the form; ln C i = α + α i ln Σw i + α ij lnσσw i + Σ β k ln y k z. lnv i + lnu i.(2.2) 1 The production frontier represents maximum output, and u i is subtracted from it. 15

16 Where; C is total cost; y k is the k-th output; w i is the i-th input price; z is the equity capital; v is measurement error term; and u is the inefficiency term. The function could further be decomposed to; ln (y it ) = x it β + v it - u it...(2.3) i= 1,2... N; t = T Where; y it is the cost per output of i-th firm in the t-th time period, x it is a K vector of value of bank variables used as explanatory associated with functional specification, β is K-vector unknown parameter to be estimated, v it are random errors assumed to be independently normally distributed, with u it s and u it being technical inefficiency effects. Equation (2.3) is the reduced form of the cost function that is utilized to estimate bank profitability. Different distributions of u it s are assumed for different panels (Coelli, Rao and Battese, 1998). The model permits estimations of unbalanced panels and u i s are assumed to be exponential function of time, involving only one unknown parameter. Estimating bank profitability, this study also adopts a similar framework as also applied by: Wilson et.al. (2004) on European banks; Naceur (2003) on Tunisian banks; and Panayiotis et al. (2005) on Greece banks. In these studies, bank performance measurement is expressed in terms of profitability as follows; П it = α 0i + α 1 П i t α 2 П i t-2 + α 3 g t-1 + β i X it + γmacro + u it...(2.4) 16

17 Where П is the profitability variable and X i = other bank variable indicators, and g it is the growth rate variable given by logarithmic value of bank size (proxy by total bank assets). Empirical studies identify average asset (ROAA) and net interest margin (NIM) as common possible choices for measuring bank profitability, though at times return on average equity have been used. The former, however are common features in this reviewed literature for this study. In the same way, were adopted for this study Panel specification for bank profitability Panel specification adopts a cost efficiency functional framework of equation (2.3) and expressed as follows; П it = c + Σβ i X it +ε it...(2.5) ε it = η i + λ t +v it ;which is a two way error correction component. Where П it is profitability of bank i at time t, with i =1, N; t = 1, T; c is a constant term; η i is the unobservable bank specific effects; and λ t is the time-specific effects and v it is the remainder error term assumed to be white noise stochastic error term, α is a constant and β is a (Kx1) vector of the coefficients of K explanatory variables as bank indicators. Using log linear transformation, equation (2.5) decomposes to; lnп it = c + β i lnx it + ε it (2.6) 17

18 The explanatory variables X it s are grouped into bank specific; industry specific and macroeconomic variables. Thus, a general specification of the profit function becomes; ln П it = c + β i ln J X it + β i ln X I it + β i ln X it M + ε it ) Where; X it s with subscripts J, I and M, denote bank specific, industry-specific and macroeconomic determinants, respectively. The model is represented as a two way error correction component where ε it is given as: ε it = η i + λ t +v it.2.8) Model specification and variables Model specification and variable identification was implemented in line with Cornett and Tehania (1992), Naceur (2003) and Panayiotis et al. (2005) classification of bank indicators. Bank indicators are classified into six categories: profitability that measures the overall performance of the bank; capital adequacy that measures the bank ability to meet regulated capital standards; credit risk that measures changes in the bank loan quality and risk; operational efficiency that measures the bank ability to generate revenue, pay, expenses and measure of employment expense; liquidity ratio that measures the changes in the bank cash position; and growth indicator that measures the bank change in assets. It is on the basis of this classification that the regression analysis of estimating bank profitability was done. 18

19 Bank profitability is the dependent variable in this study. Bank profitability can be efficiently represented by two alternative measures: return on average asset (ROAA) and net interest margin (NIM) alternative measure (IMF, 2002; Yigremachew, 2008; and Weaver, 2001). Thus, depending on data availability and consistency, these measures are applicable to study bank profitability. Return to average assets reflects the bank ability to generate profits from bank assets although it may be biased due to off-balance sheet activities (Panayiotis et al., 2005 and 2006). Return on assets is often referred to as the bank equity multiplier and measures financial leverage of the bank. Depending on data availability and consistency these variables have been extensively applied in measuring bank profitability. In this study, these variables were adopted for estimation and their efficiency compared in the analysis. Other bank variable indicators taken as explanatory are also explained. Capital adequacy expressed as equity to total asset ratio (eta), growth in bank deposits (log.td), operational efficiency as cost to income ratio (ctir), liquidity as net loan to total asset ratio (nlta), and growth indicator as growth in bank asset (log.ta ). To isolate the effects of bank characteristics, it is necessary to control for other factors that are used as determinants of bank profitability. In addition, the macroeconomic variables that influence bank profitability; GDP growth and inflation expectation were also included in the specification. Industry-specific variables such as bank concentration were not included due to data limitation and lack of a clear formula to estimate the variable. Even then, there is evidence that these variable representatives may have less significance to 19

20 bank s profitability. Their impact is reported largely to depend on other factors. The study therefore concentrated more on bank-specific and macroeconomic factors. Using the profitability function equation (2.7) and considering actual variable notations, the specification is given by; П it = c + β 1 lnta it + β 2 ETA it + β 3 lntd it + β 4 CTIR it + β 5 NLTA it+ β 6 lngdpa it +β 7 INFL it-1 + ε it......(2.9) Where П it : is profitability variable represented by either return to average assets (ROAA) or net interest margin (NIM), (LnTA) is growth in bank assets, (ETA) is bank equity to total assets, (lntd) growth in bank deposits, (CTIR) is cost to income ratio, (NLTA) is net loans to total assets, (lngdpa, is GDP-growth and (INFL) is inflation expectation given by current inflation. To capture profit persistence over time in the panel bank data, a dynamic specification was adopted in accordance with (Berger et al. (2000) and (Baltagi, 2001) 2. A dynamic specification includes a lagged dependent variable and is given as; П it = c + δп i,t-1 + β 1 lntta it + β 2 ETA it + β 3 lntd it + β 4 CTIR it + β 5 NLTA it +β 6 lngdpa it +β 7 INFL it-1+ ε it... (2.10) 2 Few studies have considered profit persistence in banking as indicated by (Le vonian, 1993; Roland, 1997; Eichengreen and Gibson, 2001; Goddard et al., 2004 and Gibson, 2005). In the industrial organization literature and important contribution is Geroski and Zacquemin (1988). 20

21 (П i, t-1) is the one period lagged profitability and δ is the speed of adjusting to equilibrium. A value of δ between 0 and 1 implies that profits persist, but they will eventually return to their normal (average) level. A value close to 0 means that the industry is fairly competitive (high speed adjustment), while a value of δ close to 1 implies less competitive structure (very slow adjustment) (Panayiotis et al., 2005) Determinants of commercial banks profitability and hypotheses In estimating bank profitability for SSA commercial banks, the profitability variable was considered as dependent variable. Empirical literature suggests return on average assets ROAA and net interest margin NIM, as appropriate choices for measuring bank profitability. These have been adopted in this study to provide comparative results. Bank specific variables and expected impact on profitability Bank growth indicator is given by natural logarithm of total bank assets. Boyd and Runkle (1993) established a significant inverse relationship between size and return on assets in U.S banks from 1971 to 1990 and positive relationship between financial leverage and size of banks. Berger, et al. (1987) showed that banks experience some diseconomies of scale to negatively affect performance. Goddard, et al. (2004), on five European countries, observed that the growth in bank size could positively influence bank performance. These observations suggest that the expected impact of bank size on bank profitability could be mixed. 21

22 Capital adequacy indicator measured by bank equity to total assets, refers to the amount of own funds available to support a bank business and acts as a safety net in the case of adverse selection. It could also measure the bank s ability to withstand losses. Banks with substantial capital adequacy ratio may be over cautious, passing up profitable investments opportunities. Alternatively, a declining ratio may signal capital adequacy problems. Capital is an important variable in determining bank profitability, although in the presence of capital requirements, it may proxy risk and also regulatory costs. In imperfect capital markets, well-capitalized banks may need to borrow less in order to support a given level of assets, and tend to face lower cost of funding due to lower prospective bankruptcy costs. Athanasoglou et al. (2005) and Berger. (1995) noted that in the presence of asymmetric information, a well-capitalized bank could provide a signal to the market that a better-thanaverage performance should be expected. Well-capitalized banks are, in this regard, less risky and profits should be lower because they are perceived to be safer. In this case, we would expect to observe a negative association between capital and profits. However, if regulatory capital represents a binding restriction on banks, and is perceived as a cost, we would expect a positive relationship to the extent that banks try to pass some of the regulatory cost to their customers. Profits may also lead to higher capital, if the profits earned are fully or partially reinvested. In this case, we would expect a positive causation from profits to capital. Athanasoglou, et al. (2005b) found a positive and significant effect of capital on Greek banks bank profitability, reflecting the sound financial condition of the banks. Likewise, Berger (2005) established a positive causation in both direction between capital and profitability. Hence, the expected influence of this variable could either positive or negative. 22

23 Credit risk indicator can be represented by different measurements including loans loss provision to total loans ratio as well as growth in bank deposits. Higher provisions for loan losses could signal the likelihood of possible future loan losses, though it could also indicate a timely recognition of weak loans by prudent banks. Some researchers have used loan loss provisions to measure credit risk. Loan loss provisions are part of the accounting breakdown of the revenue itself, which would, apriori, induce a significant negative correlation between the two variables. Loan loss provisions are also likely to account for realized losses rather than risk. On the other hand, deposit-to-loan ratio could also measure different levels of credit risk across countries if the respective practices on income verification and collaterals are different. Al-Haschimi (2007) found a positive effect of credit risk on Sub-Saharan Africa s net interest margins. With perfect capital markets and no bankruptcy costs, the capital structure (how assets are financed) does not matter and value can only be generated by the assets. However, with asymmetric information and bankruptcy costs, the specific way in which assets are funded could create value. So the expected impact of this variable to bank profitability could be mixed. Operational efficiency indicator which is also referred to as expenses by management is given as cost to income ratio. The higher this ratio, the less efficient and bank could adversely be affected in return on assets, depending on the degree of competition in the market. Al-Haschimi (2007) showed that operating inefficiencies appear to be the main determinants of high bank spreads in SSA economies. Brock and Rojas Suarez (2000) also established that administrative and other operating costs contribute to the prevalence of 23

24 high spreads in Latin American countries. Some other studies (Bourke, 1989; and Molyneux and Thornton, 1992) revealed a positive relationship between better quality management and profitability in European banks. This variable could therefore have a positive or negative impact on bank profitability, positive with better quality management at reduced costs, negative at higher inefficiency levels at higher costs. Liquidity risk indicator is measured by bank net loans to total assets or a percentage of assets that comprise the loan portfolio. High ratios may be an indicative of better bank performance because of possible increases in interest income. However, very high ratios could also reduce liquidity and increase the number of marginal borrowers that default. This is also considered as bank activity mix and also an important proxy for the overall level of risk undertaken by banks to the extent that different sources of income are characterized by different credit risk and volatility. Demirgüç-Kunt and Huizinga (1998) study of banks in 80 countries found that those with relatively high non-interest earning assets are, in general, less profitable. Banks that rely on deposits for their funding are also less profitable, possibly due to the required extensive branch network, and other expenses that are incurred in administering deposit accounts. Thus Again, the effect to bank profitability of this variable could be mixed. Macroeconomic variables and expected impact on bank profitability Bank performance is expected to be sensitive to macroeconomic control variables. The impact of macroeconomic variables on bank risk has recently been highlighted in the literature. GDP growth is adopted as a control for cyclical output effects, and expected to 24

25 have a positive influence on bank profitability. As GDP growth slows down, and, in particular, during recessions, credit quality deteriorates, and defaults increase, thus resulting into reduced bank returns. Demirgüç-Kunt and Huizinga (1998), and Bikker and Hu (2002) discovered a positive correlation between bank profitability and the business cycle. By employing a direct measure of business cycle, Athanasoglou, et al. (2005) found a positive, notwithstanding asymmetric, effect on bank profitability in the Greek banking industry, with the cyclical output being significant only in the upper phase of the cycle. Al-Haschimi (2007) further established that the macroeconomic environment has only limited effect on net interest margins in SSA countries. This evidence is consistent with the results of other countryspecific studies (Chirwa and Mlachila (2004) for Malawi, and Beck and Hesse (2006) for Uganda). GDP growth is therefore expected to have mixed impact on bank profitability depending on trend growth of the economy. The account for macroeconomic risk is also by controlling for inflation. It is envisaged that the extent to which inflation affects bank profitability depends on whether future movements in inflation are fully anticipated, which, in turn, depend on the ability of firms to accurately forecast future movements in the relevant control variables. An inflation rate that is fully anticipated increases profits as banks can appropriately adjust interest rates in order to increase revenues, while an unexpected change could raise costs due to imperfect interest rate adjustment. 25

26 Other studies, for example, Bourke (1989), Molyneux and Thornton (1992), Demirgüç- Kunt and Huizinga (1998), have found a positive relation between inflation and long term interest rates with bank performance. Inflation rate is approximated by the previous period s actual inflation and could positively or negatively influence bank profitability, positive due to the ability of bank management to satisfactorily, though not fully forecast the future inflation, which in turn could be incorporated into interest rate margins to achieve higher profits. The expected impact of this macro variable is therefore mixed Robustness and specification tests In the estimation of panel data, both static and dynamic specifications were checked using both fixed FE and random effects RE estimators. In addition, efficiency evaluation was alternately analyzed by including return on average assets (ROAA) and net interest margin (NIM) as dependent variables, representing bank s profitability. To check for efficiency between the feasible generalized least square FGLS and pooled least square dummy variable estimators LSDV, the Modified Wald Statistic test was applied. Further, testing for efficiency between the random effects and fixed effects estimators is by Hausman Specification test. The efficiency of the Generalized Method of Moments GMM IV was as an estimator, was also tested. The modified Wald Statistic result confirmed a rejection of the null hypothesis at 5 percent and 10 percent levels of significance; suggesting that the FGLS technique was appropriate for the study. Hausman Specification test confirmed the efficiency of the RE estimator in the measurement of bank profitability for SSA commercial banks. 26

27 Regression results revealed that the random effects (RE) technique generated more efficient results than the fixed effect (FE).This is consistent with theory that random effects estimator is expected to generate more efficient results where a lagged dependent variable is included as explanatory. Efficiency is achieved in controlling for a possible endogeinty and auto-correlation effects associated with dynamic lag models (Arrellano and Bover (1995) and Blundell and Bond (2000). The evaluation also looked at the efficiency of the GMM-IV estimator and proved the technique and found it inefficient for estimating the bank equation. The findings confirmed by Blundell and Bond (2000) that in panels of shorter time periods (T) and larger observations (N), which was the characteristic of this study, fixed effects and random effects models tend to generate more efficient results than the Generalized Method of Moments GMM-IV estimator. Given the efficiency advantage over other estimators, the results are based on the random effect estimation method. Panel unit-root and Co-integration tests were also implemented using the generalized Dickey-Fuller (DF) test using the Fisher-test that is appropriate for unbalanced panels (Baltagi (1999). It is indicated that this is an appropriate choice for testing non-stationarity in the panel as the null hypothesis. The fisher-test uses four other tests including inversechi-squared test (P), inverse normal (Z), inverse logit (L*) and modified inv.chi-squared (PM). Rejection of the null hypothesis is when the P-Values are less than the critical values at 0.01, 0.05 and 0.10, percent levels, respectively. Baltagi (1998 &1999) concluded that when panels are stationary, it so happens that they are integrated and could at least generate 27

28 at least one co-integrating equation. The specification checks also included some interaction analysis of at least two to three paired of variables to check their combined effect to bank s profitability Methodology, empirical data and analysis To construct the sample, data was drawn from financial statements of individual banks provided in the Bank-Scope-Database. The Bank-Scope Database is a collection of data of balance sheets, income statements and other relevant financial accounts of several banks in the World. The data base was accessed through Bank of Uganda (BoU). To ensure consistency, only data for commercial banks in the unconsolidated format was used. The period of study is 1999 to Mathieson and Roldos (2001) indicated three important characteristics of the Bank-Scopedatabase. First, its comprehensive coverage as Bank Scope data on banks accounts for around 90 percent of total bank assets in each country. Second, comparability, the datacollection process is based on separate data templates for each country to accommodate different reporting and accounting standards. Bank-Scope adjusts the collected data for country variation and presents them in a so-called global format. It is a globally standardised form for presenting bank data. Thus, Bank-Scope Databases are comparable across banks and across countries and allows cross-country comparisons (Claessens, Demirguc-Kunt and Huizinga, 2001). Third, Bank-Scope Databases provides information for individual banks, which are usually not available from other sources. Other data sources included International Monetary Fund- Financial Statistics. 28

29 Data was generated from 42 countries and 224 commercial banks with at least two years of operation between 1999 to In total, there were 1316 observations. In order to account for time trend in the data set, a dynamic specification with a lagged dependent variable used as explanatory. The variables considered in the specification include: bank asset growth in assets, capital adequacy, credit risk, operating efficiency, liquidity ratio. Macroeconomic variables that measure the influence on bank profitability also to augment the explanatory variables. These include growth in GDP and inflation. Data was downloaded in Microsoft Office, arranged in panel sets and analyzed using STATA- 10 and II, respectively. In order to understand the variability in banks profitability across the SSA, the countries were further categorized in low income GDP-per-capita below 750 US dollars, medium income GDP-Per-capita between 750 to 1500 US dollars and higher income with GDP percapita of above 1500 US dollars per year. In the analysis, although an attempt was made to uncover the interaction effects of some related pair of variables, the results were not efficient and therefore dropped in the analyses. 2.7 Discussion of Results Introduction This section provides a discussion of the quantitative results for the study. They focus on data characteristics and regression relationship of the bank level as well as macroeconomic factors to banks profitability for sampled banks from 42 countries. While the first part gives a brief summary of data characteristics in terms of unit-root test, descriptive statistics and correlation relationship in form of tables, the second part presents a detailed presentation of the regression results. Results showed that bank-specific as well as 29

30 macroeconomic variables have significant influence to the bank s profitability Data characteristics a) Descriptive statistics of the variables Table 2.1 presents the descriptive statistics of the variables utilized in this study. The results confirm the adequacy of the data used in estimating commercial banks profitability, all ranging above 1000 observations. Looking at the minimum, mean and maximum values, generally, the statistics indicate a wide variation in both the bank-specific and macro determinants of profitability of banks across the sub-regional groupings of countries within the SSA. This had an important implication to the approach used in estimating commercial banks profitability in the SSA countries, by dividing the countries in low-income, medium income and high-income categories, using panel data technique in the empirical analysis to correct for possible variations. On the average, poor performance of the bank indicators variables, with a wider variation across counties in the sub-region. Poorest performing countries lie in the low income category amongst which experienced political instability including Burundi, Gambia, among others. This category constituted about 73 percent of the banks. Better performing countries lie in the medium and high income categories and includes Botswana, Cameroon, and Madagascar among others, which implemented successful financial sector reforms during the 1990s (IMF, 2005). 30

31 Table 2.1: Descriptive statistics of the variables Variable Obs. Mean Min Max Return on average asset Net interest margin banks assets Bank deposits Operational efficiency Capital adequacy Liquidity GDPA e e e+10 Inflation rate Source: Panel estimates: b) Correlation between variables In table 2.2, the correlation relation between variable is described. The results confirmed some level of correlation between dependent variable (return on average assest or net interest margin) and independent variables (bank assets, bank deposits, operational efficiency, capital adequacy, GDP and inflation). Overall, with the correlation relationships between the variable in the range below 0.5, it would indicate that multicollinearity was not an issue in these estimations, as no two variables were highly correlated. Table 2.2 Correlation matrix between variables Variable roaa nim lta td ctir eta nlta lgdpa inf Return on average asset Net interest margin Total bank assets Bank deposits Operational efficiency Capital adequacy Bank liquidity GDPA Inflation Source: Panel estimates: c) Panel unit-root and co-integration test To check the efficiency of the variables in the model, unit-root test for stationary and cointegration was applied. This was by the augmented Dickey-Fuller test, with Fisher-type unit-root test for unbalanced panels (Baltagi, 1998) and other econometric literature confirm this method as appropriate for unbalanced panels as it accommodates for any 31