Econ Change Restruct (2016) 49:71–93 DOI 10.1007/s10644-015-9174-6

Market power, efficiency and bank profitability: evidence from Ghana Abdul Latif Alhassan1 • Michael Lawer Tetteh2 Freeman Owusu Brobbey3

•

Received: 14 April 2015 / Accepted: 25 September 2015 / Published online: 9 October 2015  Springer Science+Business Media New York 2015

Abstract This study examines the determinants of bank profitability in Ghana within the market power, relative market power and efficient structure frameworks. Using annual data on 26 Ghanaian banks from 2003 to 2011, we employ the Herfindahl Index and concentration ratio as our proxies for market power hypothesis while efficiency scores from the data envelopment analysis is employed as a proxy for the efficient structure hypothesis. The system generalized method of moment is employed to estimate a panel data model with return on assets, return on equity and net interest margin as our proxies for bank profitability. The results of the empirical estimation reject both the market power and relative market power hypotheses in the Ghanaian banking industry. While technical efficiency is found to have a positive relationship with profitability to support the efficient structure hypothesis, a negative relationship between scale efficiency and profitability is reflected by the inability of banks to operate at the optimal scale of operations. We also document evidence on the low persistence of profit which suggests a competitive banking industry. Implications for industry regulation are discussed. Keywords Africa

Market power  Efficiency  DEA  Profitability  Banks  Ghana 

& Abdul Latif Alhassan [email protected]; [email protected] Michael Lawer Tetteh [email protected] Freeman Owusu Brobbey [email protected] 1

Graduate School of Business, University of Cape Town, Cape Town, South Africa

2

Zenith University College, Accra, Ghana

3

Methodist University College, Accra, Ghana

123

72

Econ Change Restruct (2016) 49:71–93

1 Introduction The theoretical framework that explains the linkage between market power, efficiency and profitability is grounded in the structure-conduct-performance (SCP) hypothesis of Bain (1951) and Baumol (1982) and efficient structure hypothesis of Demsetz (1973). The SCP hypothesis posits that the structure of a market influences firms pricing conduct and ultimately performance. Under the SCP framework, the market power (MP) hypothesis argues that collusion among firms with market power results in higher pricing and profitability. Applied to banking markets, the MP hypothesis suggests that banks with market power collude to charge high fees on loans and advances and non-traditional activities and lower rates on customer deposits. A variant of MP, referred to as the relative-market-power (RMP), however argues that the transmission mechanism from market structure to performance occurs through product differentiation and improved service quality which are normally associated with dominant firms and not collusive behaviour. In contrast to both the MP and RMP hypotheses, Berger (1995) also argues that bank profitability in banks can be influenced by increased efficiency. This is underpinned by the efficient structure (ES) hypothesis of Demsetz (1973). This theory argues that efficient firms enjoy lower production cost which is translated into lower pricing. This results in increased sales and higher market shares, hence high profitability. Although these theories have been applied severally to analyse determinants of profitability in developed banking markets in Europe and America (Molyneux and Thornton 1992; Lloyd-Williams and Molyneux 1994; Molyneux and Forbes 1995; Maudos 1998; Chortareas et al. 2011; Garza-Garcia 2012; Nguyen and Stewart 2013; Trujillo-Ponce 2013), evidence from developing economies appears very scanty. Unlike developed economies, banking markets in emerging countries are characterized by imperfect market conditions, high levels of concentration and less restrictive regulatory regimes. This paper therefore seeks to fill this gap in the empirical literature from the perspective of an emerging African economy. Against this background, the objective of this paper is to examine the determinants of bank profitability within the SCP and ES paradigms in Africa using Ghana as a case study. To achieve this objective, annual data on 26 banks from 2003 to 2011 is employed to estimate technical and scale efficiency scores using the data envelopment analysis (DEA) technique of Charnes et al. (1978) and Banker et al. (1984) as proxies for the ES hypothesis. The Herfindahl–Hirschman Index and market share are used to proxy for the SCP framework in market power and relative market power hypotheses. Using return on assets, return on equity and net interest margin as indicators of bank profitability, we estimate a dynamic panel model to examine the effect of market structure and efficiency on bank profitability while controlling for bank level and macroeconomic conditions. The study focuses on the Ghanaian banking market for two main reasons. First, as compared to advanced countries, the banking industry of low-and middle-income countries is characterized by high interest spread (see e.g. Tennant and Folawewo 2009). More specifically, countries in sub-Saharan Africa, Latin America and the Caribbean are bedevilled with this trend as compared to the world’s most advanced

123

Econ Change Restruct (2016) 49:71–93

73

countries and in emerging countries like Mexico, Chile and Turkey (Economic Cooperation and Development countries) (Chirwa and Mlachila,2004). This phenomenon is a sign of banking resource wastage in developing countries (Sologoub 2006) and affects both the efficiency of financial intermediation and consumer welfare. Hence, a study of this nature provides regulatory authorities with insights that will go a long way in addressing such disparity. Second, while some economic similarities in sub-Saharan African countries, the banking sectors of these countries are not the same. Chirwa and Mlachila (2004) notes that high interest spread in developing countries will linger on if the financial sector reforms fail to directly affect the market structure that the banks operate in. As it stands, the Ghanaian banking sector is still characterized by high interest spreads despite the implementation of financial reform programs which reflects some failures of the reforms. Notwithstanding, the liberalization of the Ghanaian banking sector coupled with sustained periods of political stability has led to the emergence of new indigenous banks and the arrival of multinational banks as compared to other African countries. For instance, unlike other African countries1 that encountered numerous banking sector challenges during during the implementation of financial reforms, Ghana’s implementation process has been considered as successful (Moyo et al. 2014). Such a distinct market provides a unique opportunity to empirically understand how the diversity and the current market structure have impacted on profitability and efficiency of the industry as a whole. These developments provide an interesting context to test the hypotheses on the drivers on bank profitability. The study contributes to studies on bank profitability in Ghana in three ways. First, we extend the study of Aboagye et al. (2008) on determinants of bank interest spread by including two other proxies of bank profitability which are return on assets and return on equity. This enables us to capture a growing aspect of bank revenue in non-interest income which is not captured by interest income. Second, we analyse the determinants of bank profitability within the SCP and efficient structure hypotheses. While such analysis abounds in many developed banking markets (See Berger and Hannan 1998; Molyneux and Thornton 1992; LloydWilliams and Molyneux 1994; Molyneux and Forbes 1995; Maudos 1998; Chortareas et al. 2011; Garza-Garcia 2012; Nguyen and Stewart 2013; TrujilloPonce 2013), little evidence exists in the Ghanaian case. To the author’s best knowledge, this is the first study to examine the SCP and ES hypotheses in the Ghanaian banking industry. Additionally, the dataset also provides a current update of profitability determinants in the banking industry. The final contribution of this paper is the examination of profitability persistence in the Ghanaian banking industry. This is done with the use of a dynamic panel data modelling. The persistence of profitability provides evidence on the nature of competition in the banking industry. As far as we are aware, this is the first study to conduct such analysis for the banking industry in Ghana. The rest of the paper is organized as follows; Sect. 2 gives a brief overview of the Ghanaian banking industry; Sect. 3 reviews empirical studies on bank profitability;

1

For example Nigeria.

123

74

Econ Change Restruct (2016) 49:71–93

Sect. 4 discusses the empirical strategy; Sect. 5 discusses the main findings whiles Sect. 6 concludes the study.

2 Overview of Ghanaian banking industry Prior to the year 2000, there were only about 12 banks in Ghana and the banking industry was dominated by Barclays bank, Standard Chartered bank and GCB bank. The periods after the year 2000 witnessed the influx of more foreign banks which deepened competition in the industry. Currently, there are about 27 Deposit Money Banks, 49 Non-Bank Financial Institutions, and 135 Rural and Community Banks (RCBs). The banking industry continues to be innovative, profitable, liquid, and solvent despite the financial crisis the world is battling with. The industry, in recent times, has witnessed developments in the payment systems, which include an extension of the Cheque Codeline Clearing with the cheque truncation system nationwide as well as an upgrade of the Ghana Interbank Settlement System. In 2012, the financial sector also experienced some changes in the competitive environment as a result of mergers and acquisitions. Access Bank Ghana with a market share of 1.20 % acquired Intercontinental Bank of Ghana, which had a market share of 3.20 % while The Trust Bank Limited with a market share of 2.80 % was acquired by Ecobank Ghana, with a market share of 9.00 %. In addition, the minimum capital requirement has been revised to 120 million Ghana cedis from 60 million Ghana cedis after 2013 for new entrants. A summary of the bank assets and liabilities from 2003 to 2011 are presented in Table 1. Over the study period, the total assets of the industry increased from 25.2 % to attain a maximum in the year 2007 with a growth rate of 50.4 % but years after, witnessed the total assets decreasing to 26.8 % in 2011. On the whole, the industry recorded an average growth rate of 33 % which shows increase investment in bank assets such as loans and advances and other tangible assets. 47 % of the total assets consists of loans and advances by the banks which imply that banks in the country do well in advancing credit to deficit units.

Table 1 Bank total assets, deposits and loans and advances

All values in millions of Ghana cedis. Source: Bank of Ghana financial stability reports (2003–2011)

123

Year

Total assets

Deposits

Loans and advances

2003

2300

1650

890

2004

2880

2120

1130

2005

3679

2570

1580

2006

5184

3630

2520

2007

7796

4914

4147

2008

10,692

6949

5967

2009

14,043

8971

6921

2010

17,398

11,811

7995

2011

22,059

15,991

9352

Econ Change Restruct (2016) 49:71–93

75

Deposit also grew over the study period but attained a maximum growth rate of 41.4 % in 2008 during the peak of the credit crunch in the world. The periods after 2008 witnessed decrease in the growth rate. Over the period, the industry disbursed about 69.1 % of its deposit as loans which is not in violation of the Central bank’s reserve requirement of 9 % and loanable fund of 91 %. Loans and advances increased 64.6 % in 2007 but fell steadily to 17 % in 2011. Over the study period, the industry has recorded a decrease in loans and advances and this may be attributed to increase in interest rates triggered by consistent increase in policy rate making the cost of borrowing expensive thereby discouraging borrowing (Fig. 1). Table 2 reports HHI and concentration ratio (CR) of the Ghanaian banking system from 2003 to 2011. The highest concentration ratio uses customer loans measure involving five banks (denoted CR5) which was 71.46 % in 2003 indicating that these five banks dominated the industry in 2003. By 2011 CR5 has fallen drastically to 38.15 %. The corresponding five-bank concentration ratio measured using total assets also fell from 69.5 % in 2003 to 44.27 % by the end of 2011. Similar substantial declines in concentration ratio are also observed for the three-bank measures based on customer loans (from 53.24 % in 2003 to 25.05 % in 2011) and total assets (from 49.26 % in 2003 to 30.35 % in 2011). Across all periods, the concentration ratios declined throughout the study period. This coincided with increases in the number of banks in the industry. Similarly, the HHI shows a downward trend by falling from 2003 to 2011. In 2003, HHIloans, HHI-assets and HHI-deposits were 0.1228, 0.1141 and 0.126 respectively. During the last year of the study, these indices fell to 0.0538, 0.06 and 0.0632 respectively. On the whole, the overall downward trend of the concentration ratio and HHI suggest that the Ghanaian Banking Industry has become less concentrated over the study period. Despite the reduction in concentration, some large banks still dominate the whole banking system. 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00%

2004

2005

2006

Asset growth

2007

2008

Deposit growth

2009

2010

2011

Loan growth

Fig. 1 Growth rate in total assets, deposits and loans & advances. Bank of Ghana: financial stability reports (2003–2011)

123

76

Econ Change Restruct (2016) 49:71–93

Table 2 HHI and concentration ratios (CR) Herfindahl Index

CR5

CR3

Number of banks

Loans

Assets

Deposits

Loans

Assets

Loans

Assets

2003

0.1228

0.1141

0.126

0.7146

0.695

0.5324

0.4926

18

2004

0.1144

0.1066

0.121

0.6594

0.6559

0.5141

0.4746

18

2005

0.1039

0.0962

0.1097

0.625

0.6115

0.4684

0.4343

19

2006

0.0913

0.0871

0.099

0.5797

0.5743

0.4165

0.4129

23

2007

0.0941

0.0838

0.0862

0.574

0.556

0.4279

0.4129

23

2008

0.0863

0.0744

0.077

0.5412

0.5186

0.4046

0.3754

26

2009

0.083

0.0693

0.068

0.49

0.4946

0.3651

0.3485

27

2010

0.0597

0.06

0.0655

0.4109

0.4499

0.2908

0.3038

27

2011

0.0538

0.06

0.0632

0.3815

0.4427

0.2505

0.3035

27

Concentration ratios range from 0 to 1; HHI from 0.02 to 1; Source: Authors estimation from research data

3 Literature review The empirical relationship between market power, efficiency and profitability of financial institutions have received less attention in developing financial markets compared to developed banking markets in America and Europe. In a review of about 44 studies that examined bank profitability within the SCP and ES frameworks, Gilbert (1984) finds evidence to suggest that majority of the papers provided support for the SCP hypothesis. Later studies by Lloyd-Williams and Molyneux (1994) and Molyneux and Forbes (1995) were in line with the observations of Gilbert (1984). Earlier evidence in support of the ES hypothesis can also be found in the studies of Goldberg and Rai (1996) and Maudos (1998). However, Berger (1995) finds evidence in support of both the RMP and ES hypotheses for a sample of large banks in US. For example, Berger and Hannan (1998) analysed the relationship between market concentration and profitability of 470 banks in the US banking market. The empirical evidence suggests that the exercise of market power in a highly concentrated market results in lower deposit rates. In Europe, Molyneux and Thornton (1992) also investigated the determining factors of profitability across banking markets in 18 countries between 1986 and 1989. The authors find evidence of a positive effect of market concentration on profitability to support the MP hypothesis. In a study on a sample of Spanish banks, Lloyd-Williams and Molyneux (1994) did not find evidence in support of RMP hypothesis but rather the MP hypothesis. Molyneux and Forbes (1995) used data for banks from 18 European countries from 1986 to 1989. Their findings also supported the traditional SCP approach. Their results suggest that concentration in the European banking market lowers the cost of collusion between firms and results in higher profits for all market participants. In Malawi, Chirwa (2003) examined the market power and profitability relationship using a time series data from 1970 to 1994. From Cointegration

123

Econ Change Restruct (2016) 49:71–93

77

analysis, the author finds evidence in support of collusive behaviour among banks with market power. From the error correction terms model, the author finds a high speed of adjustment in bank profitability from a state of disequilibrium. Fu and Heffernan (2009) tested the market power and efficiency hypotheses for China during a period of important banking sector reforms. They found support for the RMP hypothesis, particularly before the periods of banking sector reforms (1992), although X-efficiency becomes more relevant in explaining bank profits as time progressed. Moreover, they did not find any support for the SCP hypothesis. Chortareas et al. (2011) also tested the market power and efficiency hypotheses for several Latin American countries for the period 1997–2005; their main findings are supportive of the efficient-structure (ES) hypothesis, disregarding any collusion in the banking sectors of Latin America. Garza-Garcia (2012) examined the determinants of bank performance in the Mexican banking sector from 2001 by testing the SCP and RMP and ES hypotheses. The results of the empirical analysis indicate that profitability of state-owned banks were mainly driven by increased market share while evidence of profitability persistence also showed a slower adjustment to industry average to suggest a less competitive banking market. Recently, Trujillo-Ponce (2013) examined profitability determinants in the Spanish banking industry covering a period from 1999 to 2009. The author found that the high bank profitability during these years is associated with a large percentage of loans in total assets, a high proportion of customer deposits, good efficiency and a low doubtful assets ratio. In addition, higher capital ratios also increased the banks’ return, but only when return on assets (ROA) is used as the profitability measure. The author found no evidence of either economies or diseconomies of scale or scope in the Spanish banking sector. The study also revealed differences in the performance of commercial and savings banks. Of all the studies2 on the Ghanaian banking industry, Aboagye et al. (2008) has analysed an aspect of bank profitability. Their study focused on the determinants of bank spread in Ghana. Hence, their study was limited to the bank interest income. The authors thus, neglected a growing sector of the banking industry in the form of non-interest income. In this study, we consider bank profitability indicators in return on assets and returns on equity in addition to the net interest spread. This enables us to examine the broader scope of bank profitability in Ghana.

4 Data and methodology 4.1 Estimating efficiency: data envelopment analysis Efficiency analysis identifies best performing decision making units (DMU) as a benchmark for inefficient ones. Firms on the production frontier are described to have achieved an efficient usage of resources to obtain an efficiency score of 1. The further away a firm moves from the frontier, the less efficient it becomes. The scores 2

Buchs and Mathiesen (2008), Aboagye et al. (2008), Biekpe (2011), Aboagye (2012), Saka et al. (2012), Alhassan et al. (2014) and Alhassan (2015).

123

78

Econ Change Restruct (2016) 49:71–93

range from 0 to 1 with best-performing firms operating on the frontier serving as role models for the inefficient firms. In this study, the data envelopment analysis (DEA) technique which applies a linear programming technique to construct a production frontier for multiple combinations of inputs and outputs was employed to estimate efficiency. The DEA allows for easy decomposition of technical efficiency into pure technical efficiency and scale efficiency. This helps in the identification of the sources of inefficiency to inform policy actions in improving efficiency. The two main orientations for the DEA are input-orientation which examines the inefficiencies in the usage of inputs holding output constant. The focus on output-orientation is output maximization with a constant input. In estimating efficiency in this study (See Farrell, 1957), we consider n banks with m different outputs produced from k different inputs. The bank specific input-oriented technical efficiency under constant returns to scale (CRS) is given by; TEðx; yÞ ¼ Min h Subject to N X

kj yi;j  yi;j

for all 8i ¼ 1; . . .; m

j¼1 N X

kj xr;j  hxr;j

for all

8i ¼ 1; . . .; k

j¼1

kj  0

8j ¼ 1; . . .; N

where h is a scalar. For a technically efficient (TE) bank, h = 1 while banks with h \ 1 are classified as inefficient. We estimate pure technical efficiency (PTE) under variable returns to scale by imposing the a convexity constraints as described P as; Nj¼1 kj ¼ 1. The ratio of TE to PTE generates the scale efficiency (SE). 4.1.1 Data, input and output variables This study employs annual bank level data extracted from the financial statements of 27 banks that operated in Ghana between 2003 and 2011. Overall, data of 26 banks were employed for the study. This represents more than 90 % of the banks operating in the country. The bank exempted was because it had only one observation for the 9 year period. All the bank level data were sourced from the Banking Supervision Department of Bank of Ghana. In the selection of the output variables, we follow the arguments of the intermediation approach which assumes that banks act as financial intermediaries who accept deposits from customers and translate them into assets in the form of loans and advances. Hence, we use customer’s deposits, fixed assets and personnel expenses as inputs variables and loans, investment assets, fees and commission income as the output variables (See Ohene-Asare and Asmild 2012; Chortareas

123

Econ Change Restruct (2016) 49:71–93

79

Table 3 Input and output variables Output variables Y1

Input variables Y2

Y3

X1

X2

X3

Mean

116,610,495

267,990,870

11,314,532

13,190,327

19,083,325

428,288,258

SD

271,067,584

342,086,294

12,367,270

17,104,956

24,166,318

600,134,784

Min

121,927

825,957

13,355

11,708

53,893

2,270,100

Max

2,204,136,732

2,065,056,490

61,150,098

94,760,008

166,951,823

4,284,732,561.

Y1 = Investment, Y2 = Loans, Y3 = Fees, X1 = Staff expenses, X2 = Fixed assets, X3 = Deposits

et al. 2011). The summary statistics of both the input and output variables are presented in Table 3. 4.2 Market structure In measuring market structure, we follow Boyd et al. (2009), Garza-Garcia (2012), Elyasiani and Wang (2012) and Nguyen and Stewart (2013) and employ Herfindahl–Hirschman index (HHI) lending (loans) as the proxy for market structure to test for the SCP hypothesis. The Hirschman Herfindahl index for lending concentration is given by; HHI ¼

N X

ms2i

i¼1

where HHI and ms2i represents the Herfindahl–Hirschman Index for loan concentration and the market share of loans for bank i in the industry. In testing the relative market power (RMP) hypothesis, we employ the market share of each banks assets to industry assets. The market share is defined below as; msi;t ¼

Total Assets of bank i in year t Total Industry assets in year t

4.3 Empirical model In line with the studies of Athanassoglou et al. (2008), and Garza-Garcia (2012), this study adopts a dynamic model to examine the SCP, RMP and ES hypotheses in the Ghanaian banking industry. We thus assume that the banking market is not in a longrun equilibrium. This helps us in eliminating any bias of assuming a long-run equilibrium. Most importantly, the dynamic specification enables us to test the persistence of bank profitability. The reduced form of the equation is presented below; pi;t ¼ b0 þ b1 pi;t1 þ b2 MPt þ b3 EFFi;t þ

n X

bi Xi;t þ ei;t

ð1Þ

i¼1

where i and t denotes bank and year respectively. p denotes bank profitability proxied as the return on assets, return on equity and net interest margin. pi,t-1

123

80

Econ Change Restruct (2016) 49:71–93

represents the lag of profitability; MP refers to the indicators of market structure proxied as the Herfindahl index (HHI) and market share (MS); EFF are efficiency scores estimated using the data envelopment analysis; Xi,t represents a vector of control variables. Equation (1) is expanded into Eq. (2) below by including bank specific variables; pi;t ¼ ai þ b1 pi;t1 þ b2 HHILt þ b3 MSi;t þ b4 XEFFi;t þ b5 SEFi;t þ b6 LOTAi;t þ b7 LLRi;t þ b8 EQRi;t þ ei;t ð2Þ where LOTA is the loans to asset ratio; LLR is the loan loss ratio; EQR is the equity ratio. We further control for the macroeconomic environment by including annual inflation and GDP growth rates to form Eq. (3) below; pi;t ¼ ai þ b1 pi;t1 þ b2 HHILt þ b3 MSi;t þ b4 XEFFi;t þ b5 SEFi;t þ b6 LOTAi;t þ b7 LLRi;t þ b8 EQRi;t þ b9 INFt þ b10 GDPt þ ei;t ð3Þ From the dynamic model specification with the lagged dependent variable as part of the explanatory variables, the use of ordinary least squares (OLS) estimation technique produces biased and inconsistent coefficients (Baltagi 2001). We thus employ dynamic panel model estimations techniques to address these shortfalls in the OLS technique. In this regard, the system generalised method of moments (GMM) estimation technique of Arellano and Bond (1991) and Blundell and Bond (1998) is employed to estimate the regression models. The systems GMM use the lagged differences of the explanatory variables as instruments instead of the level variables as the instruments. The technique also has the ability to deal with possible endogeneity and reverse causality biases. We employ the xtabond23 command in STATA12 for the estimation. In examining the validity of the system GMM estimations, we test for the validity of instruments using the Hansen J-test to assess the correct identification of the variables used in the model and the model should not have second order autocorrelation. 4.4 Hypotheses development 4.4.1 Market power hypothesis The market power hypothesis within the SCP hypothesis pioneered by Bain (1951) is of the view that the structure of a particular banking market affects the conduct and performance of such banks in the market (Ye et al. 2012). The integral conception of the SCP hypothesis is that it costs less for firms in a highly concentrated market to conspire against consumers and as such, firms that tend to have greater market power are likely to raise prices of their services thereby making super-normal profits. By implication, under such market conditions, more powerful

3

See Roodman (2009) for more on xtabond2.

123

Econ Change Restruct (2016) 49:71–93

81

firms will have domineering control of the market and make high profits at the expense of consumers due to the high concentration of the market. A significant positive relationship between the proxy for this variable and performance will therefore suggest the existence of collusive pricing in the Ghanaian banking industry. H1 : Market Power has a significant relationship with profitability 4.4.2 Relative market power hypothesis Unlike the market power hypothesis, the relative market power (RMP) hypothesis4 argues that dominant firms in large markets gains market shares and increased profitability not through collusive behavior but through perceived service quality and product differentiation as a unique selling proposition5 to charge high prices (Ye et al. 2012). This suggests that market leaders are better at the provision and delivery of better quality services to customers compared to smaller firms. The provision of banking services encompasses convenient branch locations, goodwill, higher service quality and numerous product offerings (Berger 1995; Ye et al. 2012). These services, as argued by the RMP hypothesis, are better offered by larger banks and translated into higher pricing of banking products and services. H2 : Market share has a significant relationship with profitability 4.4.3 Efficient structure hypothesis In variance with the SCP school of thought, the Efficient Structure (ES) hypothesis suggests that firms perform better than their peers simply because they are more efficient (use of less expensive inputs to produce outputs). This further culminates into gains in market share and high sales. Hence, unlike the SCP hypothesis that dwells on collusive behavior to set higher prices and achieve higher profits, the hypothesis takes the view that increased sales arising out of efficiency in production drives profitability. This results from, they’ll benefit from economies of scale which enable to them to sell at reduced prices and make higher profit. This phenomenon is supported to a large extent by the empirical findings of Shin and Kim (2011) in the Korean context while Ye et al. (2012) and Nguyen and Stewart (2013) do not find evidence to support this theory in the Chinese and Vietnamese banking sectors respectively. A significant positive relationship between the estimated efficiency scores and profitability indicators will provide support in favour of this theory.

4

The RMP hypothesis was postulated by Shepherd (1982).

5

Agudze-Tordzro et al. (2014) note that good customer service and management are key factors that lead to bank competitive advantage and profitability.

123

82

Econ Change Restruct (2016) 49:71–93

H3 : Efficiency has a significant relationship with profitability 4.4.4 Control variables

•

Bank lending

Bank lending is one of the fundamental businesses of most banks worldwide. In Ghana, the asset based of most of the banks is largely composed of loans and advances. Despite the assertion that the cost associated with managing large portfolio of loans is high (see Trujillo-Ponce 2013), interest income on loans appear to form a large portion of the profitability of banks in Ghana (Tetteh 2014). TrujilloPonce (2013) showed that banks with some substantial level of loans in their asset structure results in high profits. In this study, we expect a significant positive association between bank lending and profitability since interest on loans form substantial source of revenue for banks in the Ghanaian banking system. H4 : Bank lending has a significant positive relationship with profitability •

Risk

Per the core operational modalities of banks, they are faced with several uncertainties, the large piece of which emanates from their credit granting activities. Trujillo-Ponce (2013) found that low doubtful debts lead to higher bank activity and this has been asserted largely in the literature (see Tan and Floros 2012). This could mainly be attributed to the proportionally large provisions that banks would have to make for such negative expectations. Hence lower net profits. On the other hand, if banks pass on their inherent credit risks to credit worthy customers in the form of high interests, increasing credit risk could also result in higher bank profits. Alhassan et al. (2014) provide evidence which suggest that the banking industry is suffering from a silent form of distress due to high levels of nonperforming loans which reflects bank default risk. We therefore expect a negative influence of credit risk exposure of banks in Ghana on profitability. H5 : Bank credit risk has a significant negative relationship with profitability •

Bank capitalization

Two key concepts have often been used to explain the relationship between bank capitalization and profitability: the bankruptcy cost hypothesis and the signaling hypothesis Berger (1995). The former states that a bank with low capital risk high costs of bankruptcy and the reverse presupposes that the bank will reap high profits due to lower interest expenditure particularly on debts that have not been insured (Trujillo-Ponce 2013). The later however suggests that by increasing the capital

123

Econ Change Restruct (2016) 49:71–93

83

base of a bank, it implies that the bank is certain of future growth and profitability. To this end, then, banks with high capital base are expected to record high profits and on the side of the former, low capital formation of bank are likely to subject it to high profits. Tan and Floros (2012 p. 683) outlines the following reasons why capitalization has the tendency to increase bank profitability: well capitalized banks have good lending habit; the cost of funds for banks with high capitalization are low; well capitalized banks can grant more loans even at reasonably high risks which will then mean equally high returns for them; highly capitalized banks do not borrow more to finance their assets hence lower interest expense and higher profitability. Notwithstanding, Tan and Floros (2012) did not find any significant relationship between bank profitability and bank capitalization. AL-Omar and ALMutairi (2008) however found a significant positive relationship between profitability and bank capitalization thus supporting the above assertions by Tan and Floros (2012). Similar empirical evidence was found by Trujillo-Ponce (2013). Following the assertions of Tan and Floros (2012) and the empirical evidences provided by AL-Omar and AL-Mutairi (2008) and Trujillo-Ponce (2013), we expect a positive relationship between bank capitalization and profitability. H6 : Banks capitalization has a significant positive relationship with profitability •

Inflation

Dietrich and Wanzenried (2014) finds that low and middle income countries are mostly characterized by high rates of inflation as compared to high income countries. In relation to this, the authors further showed that inflation significantly affects the profitability of banks in lower and middle income countries in a positive way but does not have any effect on banks in higher income countries. The case for the middle and lower income countries is true particularly because high inflationary rates are surely factored into loan interest payments for customers. Thus if the bank forecasts a high inflationary trend during the credit assessment and granting process, the perceived high future rise in the inflation rate will be factored into the loan pricing model by the bank hence resulting in a positive impact on bank profitability (Perry 1992). Trujillo-Ponce (2013), Tan and Floros (2012) found similar empirical evidence of the positive significant relationship between inflation and bank profitability. We therefore expect rate of inflation to positively influence bank profitability in Ghana (Table 4). H7 : Inflation has a significant positive relationship with profitability •

Economic growth

The contribution of banks to the socio-economic development of any country cannot be downplayed since bad prevailing economic conditions have the tendency to affect businesses and households which in turn has the likelihood to impact negatively on credit portfolio performance and hence the profitability of banks. In

123

84

Econ Change Restruct (2016) 49:71–93

Table 4 Variables description and expected signs Variables

Symbols

Definition

ROA

Return on asset

ROE

Return on equity

NIM

Net interest margin

Signs

Dependent variables Profitability

Independent variables Market structure

HHIL

Herfindahl–Hirschman Index of lending

?

MS

Market share of total assets

?

XEFF

Technical efficiency

?

SEFF

Scale efficiency

?

Bank lending

LOTA

Loans to total assets ratio

?/–

Bank risk

NPL

Loan loss provisions to gross loans

–

Bank capitalization

EQR

Equity to total assets ratio

?/–

Inflation

INF

Inflation rate

–

Economic growth

GDP

GDP growth rate

?

Efficiency

Ghana, the role played by the financial sector and the reversal effect of the economy on banking operations cannot be overemphasized. Athanasoglou et al. (2008) show empirically that economic well-being of a country measured by its GDP growth has a significant positive influence on bank profitability. Similar evidence was shown by Dietrich and Wanzenried (2014) for middle and high-income countries and TrujilloPonce (2013) for Spain. Following that Ghana falls within the middle-income bracket, we expect that the prosperity or otherwise of the economy will have a direct impact on the profitability of banks. H8 : Economic growth has a significant positive relationship with profitability

5 Empirical results Table 5 presents the estimated efficiency scores in technical efficiency (TE), pure technical efficiency (PTE) and scale efficiency (SE) for the study period (2003–2011). We observe relative declines in TE, PTE and SE in the later parts of the study period notable from 2007 to 2011 after improvements between 2003 and 2004 except for SE which declines over the same period. In a similar vein, we can observe decreases in TE and PTE scores after 2004. Specifically, it can be noticed that TE reached its peak in 2006 by recording 91.75 % before taking a downward trajectory till 2011. PTE also showed an impressive upward rise from 85.32 % from the beginning of the study period (2003) to 97.28 % in 2004 before falling slightly to 96.83 in 2005 and then again shot up to 97.21 in the following year. It fell from 2006 until the end of the study period (2011). Unlike TE and PTE, SE has generally declined over time from its peak of 95.36 % in 2003. Though there

123

Econ Change Restruct (2016) 49:71–93 Table 5 Bank efficiency scores for TE, PTE and SE

TE Technical efficiency, PTE pure technical efficiency, SE scale efficiency

Years

TE (%)

PTE (%)

SE (%)

2003

81.41

85.32

95.36

2004

90.24

97.28

92.80

2005

90.22

96.83

93.13

2006

91.75

97.21

94.44

2007

90.51

96.18

94.26

2008

85.63

96.11

89.21

2009

84.57

92.81

90.99

2010

83.82

91.85

91.47

2011

81.41

90.65

90.35

Average

86.62

93.80

92.45

Table 6 Descriptive statistics

ROA Return on assets, ROE return on equity, NIM net interest margin, HHIL Herfindahl–Hirschman Index for lending, MS market share in assets, XEFF technical efficiency, SEFF scale efficiency, LOTA loans to total assets ratio, LLR loan loss ratio, EQR equity to assets ratio, INF inflation rate, GDP GDP growth rate

85

Mean

SD

Min

Max

N

ROA

0.022

0.046

-0.159

0.318

205

ROE

0.177

0.290

-0.870

1.628

205

NIM

0.087

0.143

0.000

1.308

205

HHIL

0.087

0.021

0.054

0.123

205

MS

0.044

0.042

0.001

0.192

205

TE

0.863

0.166

0.348

1.000

180

SE

0.922

0.126

0.446

1.000

180

LOTA

0.404

0.142

0.040

0.704

205

LLR

0.057

0.171

-0.004

1.541

205

EQR

0.149

0.120

0.030

0.870

205

INF

0.143

0.051

0.087

0.267

205

GDP

0.073

0.030

0.040

0.144

205

appear to have been some recovery between 2004 and 2006, the rest of the years for bank SE have been inconsistent. Overall, the technical inefficiency of 13.38 % is attributable to scale inefficiency of 7.55 % compared to pure technical inefficiency of 6.20 %. The descriptive statistics for the regression variables are presented in Table 6. This includes the three different measures of profitability (ROA, ROE, and NIM) and all other independent variables in the specified regression models. Over the study period, the banks recorded an average of 0.022, 0.177 and 0.087 for ROA, ROE and NIM respectively. It is also observed from the minimum values that at least one of the banks under study recorded negative ROA and ROE (losses) while the worst for NIM was a breakeven as shown by the zero (0.000) net interest margin over the study period. With maximum values of ROE and NIM above 1, these give indication of outliers6 in these variables. On the market structure variables, HHIL 6

Formal test to for the existence of outliers conducted supported the observation. The concerned variables were winsorized to produce normalised values.

123

86

Econ Change Restruct (2016) 49:71–93

for the industry averaged 0.087 with a minimum of 0.054 and a maximum of 0.123. In terms of market share, the industry average is 0.044. As low as this may seem, there is a maximum rate of 0.192 for the entire study period. The efficiency scores show an average of 0.863 and 0.922 technical and scale efficiencies respectively. There is however a maximum efficiency of 1.000 for both efficiency indicators. LOTA, LLR and EQR give average scores of 0.404, 0.057 and 0.149 respectively. On the macroeconomic variables, we can observe an average of 14.3 % inflation between the study period and on average, the country (Ghana) recorded a GDP growth of 7.3 % with the highest growth being 14.4 % over the entire study period. In terms of multicollinearity among the variables, a cursory look at the table indicates that there is no serious multicollinearity problem at hand except for the relatively high correlation between two of the variables (i.e. INF and GDP) and HHIL with 60.3 and 70.5 % correlation coefficients respectively. Though the correlation between HHIL and GDP (70.5 %) may appear to be quite high, it’s not high enough to warrant its elimination. Moreover, its elimination or removal is not expected to cause any misspecification of the model. Both economic variables have however been included in the model estimation as their high correlation with HHIL could also give an indication of the important role played by macroeconomic indicators in determining bank profitability. One other interesting observation from the table is the correlation between TE and SE (69.2 %). As slightly high as this might be, the inclusion of both variables in the model is justified by the theoretical underpinning of the current study (Table 7). 5.1 Regression results Table 8 shows results from the system generalized method of moments (GMM) estimation. Joint significance of the estimates was assessed using the Wald v2-test. The result shows that all the coefficients in the model are = 0. A careful look at some of the estimated models show signs of serial correlation at order one AR (1)) which could lead to biased estimates. Order two (AR (2)) however shows no sign of serial correlation. HHIL and MS represent the market structure variables of the Ghanaian banking sector as a proxy for the market power hypothesis testing in relation to bank profitability. The XEFF and SEFF variables will help us test the efficiency of the banks vis-a`-vis profitability (ROA, ROE or NIM). The coefficient of lagged dependent variable is used to test the persistence of profitability over the study period. If the value of the coefficients lies between 0 and 1, it indicates low persistence of profitability and high convergence to normal profits arising out of competition. However, a coefficient closer to 1 reflects high persistence and slow convergence to normal profits. This indicates a less competitive industry. From the estimation results in Table 8, the coefficients of the lagged dependent variable ranges between a minimum of 0.188 and a maximum of 0.433. This also indicates a low persistence of profitability and reflects a competitive banking industry in Ghana. Garza-Garcia (2012) finds a high persistence of profitability in the Mexican banking industry to reflect a less competitive banking market.

123

-0.066

0.064

-0.236***

-0.189***

0.168**

-0.122

-0.026

1

MS

-0.095

-0.075

0.126*

0.043

0.076

0.692***

1

TE

-0.074

0.037

0.0174

0.051

0.1028

1

SE

0.062

-0.117*

-0.396***

-0.310***

1

LOTA

0.095

-0.129*

0.322***

1

LLR

0.088

-0.036

1

EQR

-0.563****

1

INF

1

GDP

HHIL Herfindahl–Hirschman Index for lending, MS market share in assets, XEFF technical efficiency, SEFF scale efficiency, LOTA loans to total assets ratio, LLR loan loss ratio, EQR equity to assets ratio, INF inflation rate, GDP GDP growth rate

***, ** and * denotes significance at 1, 5 and 10 %

0.603***

-0.705***

INF

-0.150**

EQR

GDP

-0.156**

SE

LLR

0.101

TE

-0.140**

0.096

MS

LOTA

1

0.134*

HHIL

HHIL

Table 7 Pearson correlation matrix

Econ Change Restruct (2016) 49:71–93 87

123

123

0.1364***

0.0706***

LOTA

LLR

EQR

0.134

0.074***

0.147***

0.035***

-0.064***

0.047*

-0.052

224.13*** -1.42(0.155) 12.52(0.326)

-1.52(0.128)

-1.52(0.131)

12.60(0.479)

22

1.18

26

156

Wald v2

AR(1): (p value)

AR(2): (p value)

Hansen J: (p value)

Number of Instruments

Instruments Ratio

Banks

Observations

0.426***

156

26

1.18

22

26(0.017)

-1.29(0.198)

-2.48(0.013)

317.18***

-0.493***

1.098***

0.065

-0.583***

0.373*

-0.251

-0.779

0.134*

156

26

1.18

22

10.34(0.5)

-0.97(0.334)

-2.43(0.015)

97.91***

3.012***

0.922

-0.286

1.180***

0.025

-0.715**

0.454*

-0.183

0.608

0.208*

-0.006

Coef.

156

26

1.18

22

16.54(0.221)

-0.24(0.807)

-2.03(0.042)

985.61***

0.235***

0.798***

0.058**

-0.142**

0.103*

-0.064

-0.061

0.307***

0.014

Coef.

NIM

156

26

1.18

22

14.54(0.204)

-0.07(0.943)

-2.05(0.04)

1317.15***

0.076

-0.162

0.261***

0.800***

0.086**

-0.227**

0.167**

0.046

0.253

0.433***

-0.004

Coef.

ROA Return on assets, ROE return on equity, NIM net interest margin, HHIL Herfindahl–Hirschman Index for lending, MS market share in assets, XEFF technical efficiency, SEFF scale efficiency, LOTA loans to total assets ratio, LLR loan loss ratio, EQR equity to assets ratio, INF inflation rate, GDP GDP growth rate

AR(1) and AR(2) are first and second order autocorrelation tests respectively, Hansen J = Test of over-identifying restrictions. ***; ** and * denotes significance at 1, 5 and 10 % respectively

156

26

1.18

22

-1.71(0.087)

0.392***

184.48***

GDP

0.153*

0.0387***

SEFF

-0.049 -0.188**

INF

0.0386*

-0.0621***

XEFF

-0.2147***

-0.0207

HHIL

MS

0.0363***

-0.2597***

Constant

L.DEP

Coef.

Coef.

Coef.

ROE

ROA

Table 8 Two-step system generalized method of moments estimation

88 Econ Change Restruct (2016) 49:71–93

Econ Change Restruct (2016) 49:71–93

89

The results show a negative statistical significance between HHIL and ROA at the 1 % significance level. However, the relationships between ROE and NIM and Herfindahl index are not significant. This indicates a rejection of the SCP hypothesis and any possible collusion among market leaders in the Ghanaian banking industry. This relationship is explained by the persistence of profitability results. The ability of banks to charge high prices rest on collusion in concentrated markets. Market share (MS) has no significant relationship with all indicators of profitability which suggests a rejection of RMP hypothesis. This is consistent with evidence in Vietnamese banking industry by Nguyen and Stewart (2013). For the ES hypothesis, we find technical efficiency (XEFF) to have a positive significant relationship with ROA and ROE at the 10-percent significance level and NIM at 5-percent significance level (in the model with inflation and GDP). This implies that banks that maximize the use of their inputs in deposits and labor are able to lower unit costs and thereby earn more economic profits. Against expectations, the results indicate a negative significant relationship between scale efficiency (SEFF) and the bank profitability indicators (ROA, ROE and NIM) at the 1-percent and 5-percent significant levels, contrary to the findings of Shin and Kim (2011) and Garza-Garcia (2012). This suggests that scale efficient banks are less profitable. This could be attributed to the high scale inefficiency in the banking market. The diseconomies of scale in the form of high monitoring cost associated with the large scale operations off-set technical efficiency gains to increases production cost. Ye et al. (2012) found similar result in the Chinese banking market. Regarding the control variables, our results show that LOTA which is a measure of liquidity risk has a significant positive relationship with ROA and NIM at the 1 and 5 % significant levels respectively for both models (with and without economic variables). This indicates that increasing the loan origination functions of banks translate into higher interest revenue and high profitability. The indicator for bank risk, LLR is significant and positive across all the models giving an indication that higher credit risk translates into higher bank profitability. One reason we can assign to this would be the practice by banks to transfer costs of non-performing loans to customers in the form of higher interest margins. Bank equity (EQR) is positive and significantly related to ROA and NIM at 1-percent significance. The positive significance under ROA and NIM suggests that higher bank capitalization reduces cost of funds to undertake profitable investment projects. The negative relationship between EQR and ROE provides an indication that Ghanaian banks have high funding cost when it comes to equity financing thereby reducing their profitability. The significant macroeconomic variables are all positively related to ROA and ROE. Thus the profitability of the banks in Ghana is affected positively by the Ghanaian macroeconomic environment and as well follows the business cycle. 5.2 Robustness analysis In order to provide robustness to our basic GMM estimations, we consider either the random effects or fixed effects estimations. The choice of using either the fixed or random effects (estimation procedures) depends on the assumptions underlying the relationship between constant terms and error terms. Results of the random effects

123

90

Econ Change Restruct (2016) 49:71–93

Table 9 Random effects estimation ROA Coef.

ROE Coef.

Coef.

NIM Coef.

Coef.

Coef.

Constant

0.016

-0.034

0.524***

0.291

0.077

0.051

HHIL

-0.112

0.139

-1.674*

-0.057

-0.125

0.352

MS

0.03

0.022

-0.07

-0.144

-0.288

-0.301

XEFF

0.063***

0.068***

0.340**

0.351**

0.119**

0.102*

SEFF

-0.075**

-0.079***

-0.464**

-0.472**

-0.142**

-0.130**

LOTA

0.011

0.014

-0.179

-0.159

-0.051

-0.044

NPL

1.004***

1.016***

5.257***

5.228***

3.562***

3.529***

EQR

0.073***

0.078***

-0.491**

-0.454**

0.205***

0.218***

INF

0.031

-0.156

-0.192

GDP

0.306**

1.401

0.172

Wald v2

82.27***

89.44***

41.14***

43.69***

213.66***

217.85***

R-squared

0.324

0.3448

0.1938

0.2061

0.5704

0.576

Hausman

8.96

10.79

7.04

6.72

12.81

10.07

Prob [ v2

0.2559

0.2905

0.4251

0.667

0.0768

0.3451

Banks

180

180

180

180

180

180

Observations

26

26

26

26

26

26

***; ** and * denotes significance at 1, 5 and 10 % respectively ROA Return on assets, ROE return on equity, NIM net interest margin, HHIL Herfindahl–Hirschman Index for lending, MS market share in assets, XEFF technical efficiency, SEFF scale efficiency, LOTA loans to total assets ratio, LLR loan loss ratio, EQR equity to assets ratio, INF inflation rate, GDP GDP growth rate

estimation is presented in Table 9 following the appropriate econometric test (Hausman test) as to which of the effects should be modeled. As indicated, the random effects estimation was carried out as a remedial measure to curtail the possibility of unobserved heterogeneity bias in our estimation (Gounder and Sharma 2012). The RE estimations also reject both the SCP and RMP hypotheses since both HHIL and MSA exhibit no significant relationship with the proxies of bank profitability. Similar to the system GMM estimations, XEFF is found to have positive and significant relationship with all three proxies for bank profitability. SEFF is negatively related to ROA, ROE and NIM as found in the basic estimations in Table 8. The relationship between the control variables and bank profitability were also similar to that of the system GMM estimations. Base on this analysis, it is evident that the efficient structure hypothesis holds in part in the Ghanaian banking industry while no evidence is found for collusive behavior.

6 Summary and conclusion This paper examines the determinants of bank profitability within the market power, relative market power and efficient structure hypotheses. Using annual data on 26 banks from 2003 to 2011, we estimate the Herfindahl–Hirschman Index of Lending

123

Econ Change Restruct (2016) 49:71–93

91

(HHIL) and market share (MS) as proxies for the MP and RMP hypotheses while technical and scale efficiency scores are estimated from data envelopment analysis (DEA) technique to proxy for the ES hypothesis. From the efficiency estimates from the DEA, we find the average inefficiency of 13.38 % to be attributable to scale inefficiency. This suggests that Ghanaian banks are operating at sub-optimal scales. The relationship between these variables and three indicators of profitability in return on assets, return on equity and net interest margins are examined with both static and dynamic panel regression framework. The results of the system GMM estimations suggest that the existence of low persistence of profitability in the Ghanaian banking industry. Additionally, we find insignificant relationship between bank profitability and both HHIL and MSA. Hence, we reject the SCP and RMP hypotheses. This suggests the absence of collusive behavior in the banking market and further support the evidence on low persistence of profitability, which reflects a competitive industry. We also find a positive effect of technical efficiency on profitability while scale efficiency reduces bank profitability. This partly provides support for the ES hypothesis in the Ghanaian banking industry. The overall implication of our study is that banks in Ghana have been able to improve their technical efficiencies over the years but have not been able to do same with their scale efficiencies. One other implication is that since market power does not necessarily lead to significant profits to the banks, it’s about time they focus on greater efficiency as the market appear to be continuously concentrated with lending. To bring robustness to our findings, we suggest that similar studies be replicated in banking markets in other African countries since the recommendations of this paper may not be applicable in those contexts due to differences in the structure of banking markets and regulatory regimes. Acknowledgments The authors acknowledge the constructive comments of an anonymous reviewer which improved an earlier draft of the manuscript. The normal caveats apply.

References Aboagye AQQ (2012) Bank concentration and economic costs of deposit mobilization and credit extension in Ghana. J Dev Areas 46(2):351–370 Aboagye AQQ, Akoena SK, Antwi-Asare TO, Gockel FA (2008) Explaining interest rate spreads in Ghana. Afr Dev Rev 20(3):378–399 Agudze-Tordzro K, Buame S, Narteh B (2014) Customer retention strategies: a study of Ghanaian banks. Eur J Busin Manag 6(15):31–39 Alhassan AL (2015) Explaining bank productivity in Ghana. Manag Decis Econ. doi:10.1002/mde.2748 Alhassan AL, Kyereboah-Coleman A, Andoh C (2014) Asset quality in a crisis period: an empirical examination of Ghanaian banks. Rev Dev Finance 4(1):50–62 AL-Omar H, AL-Mutairi A (2008) Bank-specific determinants of profitability: the case of Kuwait. J Econ Adm Sci 24(2):20–34 Arellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev Econ Stud 58(2):277–297 Athanasoglou P, Brissimis S, Delis M (2008) Bank-specific, industry-specific and macroeconomic determinants of bank profitability. J Int Financ Mark Inst Money 18(2):121–136

123

92

Econ Change Restruct (2016) 49:71–93

Bain JS (1951) Relation of profit-rate to industry concentration: American manufacturing, 1936–1940. Q J Econ 65:293–324 Baltagi B (2001) Econometrics analysis of panel data. Wiley, West Sussex Banker RD, Charnes A, Cooper WW (1984) Some models for estimating technical and scale inefficiencies in data envelopment analysis. Manag Sci 9(30):1078–1092 Baumol WJ (1982) Contestable markets: an uprising in the theory of industry structure. Am Econ Rev 72(1):1–15 Berger AN (1995) The profit-relationship in banking–tests of market-power and efficient-market hypotheses. J Money Credit Bank 27:405–431 Berger AN, Hannan TH (1998) The efficiency cost of market power in the banking industry: a test of the ‘‘quiet life’’ and related hypotheses. Rev Econ Stat 80:454–465 Biekpe N (2011) The competitiveness of commercial banks in Ghana. Afr Dev Rev 23(1):75–87 Blundell RW, Bond SR (1998) Initial conditions and moment restrictions in dynamic panel data models. J Appl Econom 87:115–143. Boyd JH, De Nicolo G, Jalal AM (2009) Bank competition, risk and asset allocations IMF working paper 09(143). Buchs T, Mathiesen J (2008) Banking competition and efficiency in Ghana. Economy of Ghana, 1st edn. Boydell & Brewer, Woodbridge, pp 173–194 Charnes A, Cooper WW, Rhoades E (1978) Measuring the efficiency of decision making units. Eur J Oper Res 2:429–444 Chirwa EW (2003) Determinants of commercial banks’ profitability in Malawi: a cointegration approach. Appl Financ Econ 13(8):565–571. doi:10.1080/0960310022000020933 Chirwa EW, Mlachila M (2004) Financial reforms and interest rate spreads in the commercial banking system in Malawi. IMF Staff Pap 51:96–122 Chortareas G, Garza-Garcia JG, Girardone C (2011) Banking sector performance in Latin America: market power versus efficiency. Rev Dev Econ 15:307–325 Demsetz H (1973) Industry structure, market rivalry, and public policy. J Law Econ 16:1–9 Dietrich A, Wanzenried G (2014) The determinants of commercial banking profitability in low-, middle-, and high-income countries. Q Rev Econ Finance 54(3):337–354 Elyasiani E, Wang Y (2012) Bank holding company diversification and production efficiency. Appl Financ Econ 22(17):1409–1428 Farrell MJ (1957) The measurement of productive efficiency. J R Stat Soc 120(3):253–290 Fu XM, Heffernan S (2009) The effects of reform on China’s bank structure and performance. J Bank Finance 33(1):39–52 Garza-Garcia JG (2012) Does market power influence bank profits in Mexico? A study on market power and efficiency. Appl Financ Econ 22(1):21–32 Gilbert RA (1984) Bank market structure and competition: a survey. J Money Credit Bank 16:617–645 Goldberg L, Rai A (1996) The structure–performance relationship for European banking. J Bank Finance 20:745–771 Gounder N, Sharma P (2012) Determinants of bank net interest margins in Fiji, a small island developing state. Appl Financ Econ 22(19):1647–1654 Lloyd-Williams DM, Molyneux P (1994) Market structure and performance in Spanish banking. J Bank Finance 18:433–443 Maudos J (1998) Market structure and performance in Spanish banking using a direct measure of efficiency. Appl Financ Econ 8:191–200 Molyneux P, Forbes W (1995) Market structure and performance in European banking. Appl Econ 27:155–159 Molyneux P, Thornton J (1992) Determinants of European bank profitability: a note. J Bank Finance 16:1173–1178 Moyo J, Nandwa B, Oduor J, Simpasa A (2014) Financial sector reforms, competition and banking system stability in Sub-Saharan Africa. Presented at the joint RES-SPR conference on ‘‘Macroeconomic Challenges Facing Low-Income Countries’’, Washington, DC 30–31 Jan 2014 Nguyen TN, Stewart C (2013) Concentration and efficiency in the Vietnamese banking system between 1999 and 2009: a structural model approach. J Financ Regul Compliance 21(3):268–283 Ohene-Asare K, Asmild M (2012) Banking efficiency analysis under corporate social responsibilities Int. J. Banking, Accounting and Finance 4(2):146–171. Perry P (1992) Do banks gain or lose from inflation? J Retail Bank 14:25–30

123

Econ Change Restruct (2016) 49:71–93

93

Roodman D (2009) How to do xtabond2: an introduction to difference and system GMM in Stata. Stata J 9:86–136 Saka ANA, Aboagye AQQ, Gemegah A (2012) Technical efficiency of the Ghanaian banking industry and the effects of the entry of foreign banks. J Afr Busin 13(3):232–243 Shepherd WG (1982) Economies of scale and monopoly profits. In: Craven J (ed) Industrial organization, antitrust, and public policy. Kluwer Nijhoff, Boston Shin DJ, Kim BHS (2011) Efficiency of the banking industry structure in Korea. Asian Econ J 25(4):355–373 Sologoub D (2006) The determinants of bank interest margins and profitability: case of Ukraine. www. bof.fi/bofit/seminar/bofcef06/sologub.pdf Tan Y, Floros C (2012) Bank profitability and inflation: the case of China. J Econ Stud 39(6):675–696 Tennant D, Folawewo A (2009) Macroeconomic and market determinants of interest rate spreads in lowand middle-income countries. Appl Financ Econ 19(6):489–507 Tetteh ML (2014) Local versus foreign bank performance: the case of Ghana. Banks Bank Syst 9(3):43–49 Trujillo-Ponce A (2013) What determines the profitability of banks? Evidence from Spain. Account Finance 53:561–586 Ye Q, Xu Z, Fang D (2012) Market structure, performance, and efficiency of the Chinese banking sector. Econ Change Restruct 45:337–358

123