Introduction

The study investigates the impact of competition on banks' risk-taking behaviour in Sub-Saharan Africa (SSA) countries. The effect of competition on bank risk-taking has long been a public policy debate issue over the last 20 years, especially after the recent global financial crisis (López-Penabad et al. 2021; Heimdal and Solberg 2015; Fu et al. 2014). For instance, excessive risk-taking by banks is considered by many scholars as the key factor that led to the recent global financial crisis, which forced regulators in several countries to adopt policies to reduce competition in the banking system (e.g., Mateev et al 2022; Adu 2021; González et al. 2017). This underscores the importance of examining the effect of competition on the risk-taking behaviour of banks in the SSA region.

Theoretically, economic theories provide conflicting predictions on the impact of competition on bank risk-taking. Specifically, banking competition theories provide two contrasting views concerning the effect of competition on bank risk-taking behaviour. The first theory supports the competition-fragility view. Supporters of this view maintain that intense competition in the banking sector will reduce interest income for banks and, thus, reducing banks’ profits (e.g., Allen and Gale 2004; Keeley 1990). The reduced profits will lead to an increase in bank risk-taking due to lowered franchise value (Keeley 1990). In support, other scholars assert that as competition increases in the banking sector, the economic rent from intermediation decreases substantially, leading banks to reduce screening of potential borrowers, and thus, decreasing the overall quality of banks’ loan portfolio (e.g., Marquez 2002; Chan et al. 1986). By contrast, supporters of competition-stability view suggest that bank risk-taking decreases with competition. Proponents of this theory argue that intense competition may have an effect on the cost of capital, giving entrepreneurs and firms access to lower interest rates, which tend to boost the profitability of projects (Boyd et al. 2006; Schaeck et al. 2006; Boyd and De Nicolo 2005), thereby lowering credit risks (Cuestas et al. 2020).

A systematic review of literature concerning the impact of competition on bank risk-taking shows that the results are far from conclusive, since empirical findings vary with the period and countries analysed (Cuestas et al. 2020). For example, some prior studies provide evidence in support of competition-fragility view. López-Penabad et al. (2021) sample 117 banks in 16 European countries over the period 2011–2018, and report that competition measured by Lerner index increases the risky behaviour of banks. Based on a sample of banks in 23 developed countries over the period 1999–2005, Berger et al. (2009) observe that Lerner index has positive impact on the Z-score. Evidence of this view can also be found in other previous studies (e.g., Kabir and Worthington 2017; Bushman et al. 2016; Beck et al. 2013; Soedarmono et al. 2013; Agoraki et al. 2011). Some, but a growing number of studies also document evidence in support of competition-stability view. For example, Clark et al. (2018) sample banks in 10 Commonwealth Independent States countries over the period 2005–2013 and observe a negative relationship between Lerner index and the Z-score. Similarly, Schaeck and Čihák (2010) in a sample of banks in 10 European countries and US over the period 1995–2005 report that competition has negative impact on risk-taking. Other prior studies find similar evidence (e.g., Maji and Hazarika 2018; Mustapha and Tocci 2018; Tabak et al. 2015; Fu et al. 2014; Amidu and Wolf 2013; Boyd et al. 2006). Other evidence implies that financial sector liberalization and risk-taking behaviour have a non-linear relationship (Jiménez et al. 2013; Martinez-Mierra and Repullo 2010). Evidently, the lack of consensus among studies exploring the relationship between competition and bank risk-taking poses a significant challenge for academics and policymakers (Dam and Sengupta 2022).

However, despite extensive literature search, studies that focus on whether competition can influence bank risk-taking in SSA countries remain uncommon. Most of these studies focus on international sample, whereas others are US or Europe-specific, with comparatively few studies focusing on developing countries (e.g., Haque 2019; Lassoued et al. 2016). The closest to this study is Akande et al. (2018) who examine the impact of competition on off-balance sheet risk in some selected countries in the region. Arguably, this empirical gap offers an ideal research context to make original contributions to the existing banking literature. The analysis in this paper focuses on the impact of competition and the various banks' risk-taking measures in the SSA region. The study on SSA region is motivated by two main reasons. First, there has been extensive financial sector reforms in the region over the past 30 years. The financial sector reforms have been implemented to liberalize the banking system. Noticeably, the outcome of the financial sector reforms in the region is an intense banking competition (e.g., Akande et al. 2018; Motsi et al. 2018). Meanwhile, it has been argued that intense banking competition has its attendant challenges that can cause financial instability (Akande et al. 2018). For example, the intense competition may imply that banks in the SSA region are in a better position to take higher level of risk. Hence, post-financial sector reforms investigation to establish whether competition has beneficial or detrimental effect on bank risk-taking is crucial in the SSA region.

Second, the choice of SSA region for the study emanates from the fact that relative to developed economies, the region has weak institutional framework (Kaufmann et al. 2010). In addition, the countries have highly bureaucratic and corrupt governments with low levels of ‘‘voice and accountability’’ (Kaufmann et al. 2010), as well as weak regulation (Adu 2021). Besides, implementation of Basel accords remains uneven across the region, with higher standards adopted in only a few countries (e.g., South Africa). The situation is complicated due to lack of financial safety nets in the region (Mlachila et al. 2016). Therefore, in an event of bank failures, they cannot cover 80% of deposits as evidenced in the recent banking crisis in the region (e.g., Ghana, 2018, Kenya, 2016, and Nigeria, 2009).

As mentioned above, financial sector liberalization was widely considered as a critical catalyst for boosting economic growth, and most SSA countries and other developing economies viewed it as a remedy for addressing capital accumulation, allocation, and other developmental issues (Beck and Cull 2014). Beginning in the mid-1980s, this argument fueled the implementation of major regulatory reforms in favour of financial sector liberalization in most SSA countries (Banyen 2021). Despite these and other claims to the contrary, the financial liberalization in the SSA countries has the potential to degrade individual banks' charter value and limit their ability to collect monopoly rents (Keeley 1990). To maintain their profit profiles, affected banks may either implement creative and efficient banking techniques (Berger et al. 2009; Boyd and de Nicoló 2005), or undertake risky activities (Banyen 2021; Fiordelisi and Mare 2014; Rupello 2004; Keeley 1990).

More importantly, following the recent banking crisis in the region, academics and banking practitioners maintain that due to the lack of consensus in the empirical literature, it is critical for regulators and policymakers in the SSA region seeking greater financial integration to empirically examine the impact of financial liberalization on banking sector competition, risk-taking behaviour, and overall financial stability (Adu 2021; Banyen 2021). Nevertheless, the empirical research has yet to investigate how financial integration, particularly financial liberalization and privatization of state-owned banks, influences bank risk-taking behaviour in SSA through the channel of competition (Adu 2021; Banyen 2021). With this background, the study seeks to extend, and to also make new contributions to the existing banking literature. The paper provides new empirical evidence on how intense competition through financial sector liberalization influences bank risk-taking behaviour in SSA. To the best of our knowledge, this is the first study to conduct a comprehensive investigation of this relationship in the region. Crucially, the study focuses on post-financial sector reforms in SSA which provides a unique opportunity to test the link between intense competition and bank risk-taking. The study employs both Lerner index and H-statistic as measures of bank competition. Studies in SSA region have mainly applied Lerner index as measure of competition (e.g., Akande et al. 2018). However, the H-statistic method is one of the most widely applied assessment of competition in the literature (Leon 2015). Notwithstanding its strength, H-statistic method has seen limited application in SSA studies (Fosu 2013). Furthermore, the study employs four different and common risks faced by the banks in the SSA countries. The risk-taking measures include Z-score, non-performing loans, loan-loss provision, and capital adequacy ratio. By focusing on different types of risks, the study distinctively provides impact of competition on specific risks of banks in the region. The evidence of the study shows that competition increases bank risk-taking in the SSA countries. More specifically, in the SSA countries, there is a significant, positive statistical relationship between the Lerner index and H-statistics that measure competition, and the Z-score, non-performing loans, loan-loss provision, and capital adequacy ratio that measure bank risk-taking. Overall, the evidence of the study suggests that increased competition through financial sector liberalization in the region drives banks towards greater risk-taking behaviour in support of competition-fragility view.

The rest of the paper is organized as follows: “Literature review and hypothesis development” presents the literature review and develops the hypothesis. In “Data and methodology”, the data set and the research methodology are provided. “Empirical results” provides the empirical results of the study. “Robustness checks” includes a robustness check. Finally, “Conclusion” offers the conclusion of the study.

Literature review and hypothesis development

Because of its far-reaching implications on the economy, the impact of competition on bank risk-taking is of major importance to regulators and policymakers. Through financial sector reforms, banking regulators may liberalize the banking sectors leading to an increase in competition. Thus, any findings suggesting excessive risk-taking role of intense bank competition will suggest the existence of a policy trade-off. The study therefore seeks to improve our understanding of the influence of competition on bank risk-taking.

Competition and bank risk-taking

This study is motivated by ongoing lack of consensus in the literature concerning the impact of financial sector liberalization on bank stability, coupled with the scarcity of literature on the effect of competition on bank risk-taking in the SSA context (Banyen 2021). Financial liberalization and deregulation in the banking sector have attracted the attention of several academics and policymakers on the role of competition in the banking sector (Ali et al. 2022). For instance, the banking literature provides two contrasting opinions on the relationship between competition and risk-taking (Mateev et al. 2022; Adu 2021; López-Penabad et al. 2021; Cuestas et al. 2020). First, competition-fragility view maintains that risk-taking increases with high levels of competition. This is deeply rooted in franchise value hypothesis which suggests that competition increases risk-taking because of its negative impact on banks’ franchise value (Keeley 1990). Franchise value represents the present value of the future profits that a bank is projected to make as a going concern (Demsetz et al. 1996). Profits are earnings over all other costs such as the cost of capital (Arping 2019). Franchise value has two main sources. First, banking regulators limit competition so as to offer banks access to profits. This franchise value that banks obtain due to these restrictions of competition imposed by national regulators is known as “market-related” value (Demsetz et al. 1996). The other franchise value arises from “bank-related” sources (Demsetz et al. 1996). This include gains in key areas such as operational efficiency and differences in the nature of lending (Demsetz et al. 1996). It has been suggested that, as competition increases, banks are unable to obtain steady earnings as competition causes them to reduce their rates to levels good enough to cater for their operating expenses and other costs. Accordingly, as competition increases, the present value of earnings decreases, thereby making bank failure less expensive (Keeley 1990). Thus, competition-fragility viewpoint indicates that liberalization of the financial sector encourages both international and domestic involvement, lowering bank charter value, limiting the ability to charge monopoly rents, and pushing risky behaviour to sustain profit profiles (Banyen 2021).

The implication is that banks primarily reduce their risk-taking as a way of protecting their franchise value and monopoly profits (Arping 2019). Supporters of this viewpoint maintain that restricted competition should encourage banks to protect their high “franchise value” by pursuing safety policies that contribute to reduce risk-taking (Mateev et al. 2022; Keeley 1990). However, as profit declines with increased competition, banks tend to lower margins and thus lowering discounted net value. With lowered net value, banks are more willing to take high risk (Nilsen et al. 2016). The higher level of undertaken risk may translate into lower quality of the banks’ loan portfolio (Mustafa and Toci 2018).

Another strand of banking literature has investigated the effect of competition on risk-taking from the perspective of adverse selection problem (Mustafa and Toci 2018). Adverse selection refers to a market situation which happens when a buyer and a seller have differences in market information. In banking, this stems from imperfect information in the loan market (Heimdal and Solberg 2015). For example, banks that set the same rate for all their customers may encounter risk of unfavourably selecting the least profitable borrowers. Furthermore, as competition increases in the market, there is the tendency for a declined loan applicant to make a new application for a loan in competing banks due to ‘information dispersion’ (Marquez 2002). The adverse selection problem may be mitigated through effective screening procedures implemented by banks (Mustapha and Toci 2018). However, as competition intensifies, individual banks have minimal information concerning the credit worthiness of borrowers. This prevents proper screening by banks and results in banks granting credit to borrowers with poor credit history (Marquez 2002), thereby lowering the quality of banks’ loan portfolio (Marquez, 2002; Boot and Greenbaum 1993).

Empirical research supports competition-fragility view preposition that competition increases bank risk-taking (e.g., Noman et al. 2022, 2017; Akande et al. 2018; Leroy and Lucotte 2017; Jiménez et al. 2013; Liu and Wilson 2013; Berger et al. 2009). For example, Liu and Wilson (2013) examine the effect of competition on risk-taking in a sample of 732 banks in Japan from 2000 to 2009. The authors report that Lerner index is positively related with NPLs. Tongurai and Vithessonthi (2020) examine the effect of competition on bank risk-taking in publicly listed banks in Japan during the period 1993–2016. The authors find that bank competition is positively associated with ex ante bank risk-taking, as measured by loan growth and the interest rate margin. Furthermore, Noman et al. (2017) observe that H-statistics has a positive effect on Z-score in a sample of 180 Asian banks over the period 1990–2014. Berger et al. (2009) conduct similar analysis in an international sample of 8235 banks. The banks were selected from 23 developed markets from 1999 to 2005. The authors document that Lerner index positively and significantly influences NPLs in the 23 countries. Also, Bushman et al. (2016) examine the same link in the US banking industry with a sample of 13,730 banks from 1996 to 2012, and find a positive link between Lerner index and loan-loss provision and Z-score.

In addition, Noman et al. (2022) investigate the effect of competition on bank risk-taking from 180 banks between the years 1990 and 2017 in Association of Southeast Asian Nations and observe that deposit insurance reduces bank risk in the absence of competition by reducing both credit and insolvency risk. However, the authors find that deposit insurance exacerbates bank risk-taking and weakens banking stability in a highly competitive market. They assessed risk-taking with Z-score and NPLs, whereas competition was measured by H-statistics and Lerner index. Saif-Alyousfi et al. (2020) investigate the effect of competition on bank risk-taking behaviour of banks in GCC region for the period 1998–2016. The results of their study suggest that a higher level of bank competition adds to financial fragility. Similarly, Akande et al. (2018) report that competition increases bank risk-taking in an investigation based on 440 banks in Africa from 2006 to 2015. In spite of the interest in their research to bank risk-taking studies in SSA, the risk-taking proxy employed in Akande et al. (2018) study is skewed to only one type of risk-taking in the region, which is off-balance sheet risk. This may limit the generalization of their findings.

By contrast, competition-stability view states that bank risk-taking decreases with intense competition. This view is built on “risk shifting” hypothesis which was developed by Boyd and De Nicolo (2005). They propose that as competition in the banking system intensifies, there is a corresponding reduction in the interest rate charged by banks. Boyd and De Nicolo (2005) explain that low lending rate lessens the borrowing cost of doing business which can increase entrepreneurial activity. As borrowers pay low interest rate on loans, their motivation to engage in excessive risk-taking decreases, hence making loans safer in the banking system. This view asserts that the quality of banks’ loan portfolio is mainly influenced by their borrowers’ risk-taking behaviour and not determined directly by the banks. The resulting decrease in loan default rate leads to a decline in risk-taking in the banking system. In support, Martinez-Miera and Repullo (2010) maintain that banks invest in loans and the risk associated with the loans increases with increased loan interest rate. Thus, a reduction in loan interest rates due to intense bank competition will reduce the loans’ probability of default (Mateev et al. 2022).

Additionally, proponents of this view suggest that banks that do well in establishing strong franchise value will work to protect it (Demsetz et al. 1996). Therefore, banks with high franchise value tend to reduce risk-taking propensity. For instance, banks with large franchise value minimize granting of loans to high-risk borrowers. Also, such banks have the tendency to typically diversify their credit portfolio as a means of reducing their exposure to risk, thereby reducing risk-taking (Demsetz et al. 1996). Furthermore, Mateev et al. (2022) maintain that restricted competition can encourage banks to protect their high “franchise value” by undertaking safety policies that contribute to the stability in the banking system. The authors therefore contend that based on franchise value paradigm, banks should limit their risk-taking when the banks have market power in lending. The supporters of the competition-stability view also argue that an increase in competition in banking market offers depositors with more options in terms of where to place their deposits (Mustapha and Toci 2018). With more options in the deposit markets, depositors may “penalize” excessive risk-taking banks simply by moving their deposits to banks they consider less risky and safer (Mustapha and Toci 2018).

Empirically, prior studies find evidence for competition-stability view (e.g., Arping 2019; Mustapha and Toci 2018; Sarkar and Sensarma, 2016; Tabak et al. 2015; Schaeck and Cihak 2014; Boyd et al. 2006). For example, Boyd et al. (2006) examine the relationship between competition and risk-taking in a sample of 2500 small rural banks in US. In addition, they employ a panel of 2700 banks from 134 countries over the period 1993–2004. The authors document a negative relationship between Herfindahl–Hirschman index and Z-score. Similarly, Schaeck and Cihak (2014) investigate the same link across 10 European countries by sampling 3325 banks from 1995 to 2005. The authors observe an inverse link between Boone indicator and Z-score. Furthermore, Mustapha and Toci (2018) explore the effect of competition on risk-taking in 15 European countries. Their investigation contained 1497 banks from 1999 to 2009. The authors also report an inverse link between Lerner index and loan-loss provision. Additionally, Anginer et al. (2014) investigate the effect of competition on risk-taking in a sample of 1872 publicly traded banks in 63 countries. The authors find evidence in support of competition-stability view. The researchers explain that competition can encourage banks to take more diversified risks, and hence, the banking sector will be more resilient to shocks.

To sum up, due to the financial sector reforms in SSA countries, banks in the region operate in a challenging banking environment characterized by tighter funding conditions with rising competition for deposit and loans (Stijns and Revoltella 2016). Moreover, the emergence of pan-African banking groups and microfinance institutions have intensified banking competition in the region. The increase in competition for deposits among banks may have some implications on risk-taking behaviour of banks. For instance, the increased competition can undermine prudent banking practices, particularly as regulation and supervision in the region are weak. As banks pay higher deposit rate due to competition in the region, they face higher repayment burden and can lead the banks to undertake high risky investments. Consistent with the above discussion, the study predicts a positive link between competition and bank risk-taking. To shed more light on this issue, the study tests the following hypothesis:

H1: The risk-taking behaviour of banks increases with competition, confirming the traditional competition-fragility view in the Sub-Saharan Africa (SSA) region.

Data and methodology

Data and variable description

To test the proposed hypothesis, the study constructs a sample of banks in the Sub-Saharan Africa countries. The data set covers an unbalanced panel of 220 banks from 16 SSA countries over the period 2007–2018. These countries are Botswana, Gambia, Ghana, Kenya, Lesotho, Liberia, Malawi, Mauritius, Namibia, Nigeria, Sierra Leone, South Africa, Tanzania, Uganda, Zambia, and Zimbabwe. The countries were primarily chosen, because they have a common official language which is English. In line with similar past studies (Adu 2022; Adu et al. 2022b), this helps in collection of data from the sampled banks by eliminating language barriers. The choice of the countries also partly emanates from the similar financial sector reforms implemented across the countries over the past 3 decades. Nonetheless, the 16 SSA countries have the most matured banking and capital markets in the region. For instance, the total GDP of the 16 countries stood at US$2,885.78 billion as at 2018 as compared to the GDP of the entire SSA of US$4,200.85 billion and accounted for nearly 70% of the total SSA GDP (see Table 1). Additionally, the final data set from the selected banks was adequate to estimate Lerner index and H-statistics as measures of bank competition in the SSA banking sector.

Table 1 The GDP of the countries in SSA from 2007 to 2018 (US$ billion)
Full size table

In line with prior studies in the region, this sample was considered because of data availability as well as to accommodate entry and exit of banks during the period (Adu et al. 2022b; Akande et al. 2018). Table 2 provides the list of countries and the number of banks per country that are contained in the sample. The study excluded banks with missing data or whose annual reports were not published. Furthermore, the study excluded foreign-owned banks that published their annual reports worldwide as consolidated financial statements. Also, the study sampled banks and specialized financial institutions whose nature and operations are like that of commercial banks. Apart from GDP and inflation data that were collected from World Bank and International Monetary Fund, respectively, data for this study were sourced from BankScope, and supplemented with those from annual reports, where necessary. Table A1 in the appendix provides example of websites of the banks in the SSA countries.

Table 2 Composition of the banks by countries
Full size table

For bank risk-taking, the study employs four types of measures. First, the Z-score for the sampled banks is calculated as follows:

$$Z - {\text{score}} i,t = \frac{{ROAi,t {-} EAi,t}}{ \sigma ROAi,t },$$
(1)

where ROAi,t represents return on asset for a bank i at time t, EAi,t is the equity to total assets ratio for bank i at time t, σROAi,t denotes the standard deviation of return on assets of bank i at a time t. As a robustness check, the study employs non-performing loans (NPLs), loan-loss provision (LPROV), and capital adequacy ratio (CAR), as alternative measures of bank risk-taking. In line with Chen and Lin (2016), and Eibannan (2017), NPLs are defined as bank ratio of non-performing loans to total loans in a financial year with a larger value indicating a riskier credit portfolio. Similarly, and consistent with Mustapha and Toci (2018), and Al-Khouri (2012), LPROV denotes the ratio of loan-loss provision to total loans in a financial year. Following Mustapha and Toci (2018), and Eibannan (2017), CAR represents the ratio of banks’ capital to risk weighted assets in a particular year.

Concerning bank competition, the study employs two measures. The first competition measure is based on the Lerner index developed by Lerner (1934) and used in several banking studies (e.g., Mateev et al. 2022; Delis et al. 2016; Berger et al. 2009).

Lerner index represents the mark-up of price over marginal costs (Berger et al. 2009)

$${\text{Lerner}}\,{\text{index}}_{it} = \frac{{P_{it} {-} MC_{it} }}{{P_{it} }}.$$
(2)

In Eq. 2, pit is the output price of bank i at time t and is defined as total revenue divided by total assets. Marginal cost is estimated by differentiating the translog cost function with one output (total assets) by output (Delis et al. 2016; Berger et al. 2009). Consistent with past studies (e.g., Delis et al. 2016; Berger et al. 2009), MCit is derived following translog cost function as follows:

$$InTC_{it} = \alpha_{0} + \mathop \sum \limits_{j = 1}^{2} \alpha_{1} Inw_{it}^{j} + {\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}}\mathop \sum \limits_{j = 1}^{2} \mathop \sum \limits_{k = 1}^{2} \alpha_{jk} Inw_{it}^{j} + \beta InTTA_{it} + {\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}}\beta_{2} (InTA_{it} )^{2} + \mathop \sum \limits_{k = 1}^{2} \beta_{2j} InTA_{it} Inw_{it}^{j} + \gamma_{It} T + {\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}}\gamma_{2t} T^{2} + \mathop \sum \limits_{j = 1}^{2} \gamma_{3t} TInw_{it}^{j} + \gamma_{4t} TInTA_{it} + \varepsilon_{i} ,$$
(3)

where:

TCi = the bank’s total costs

TAi = the total assets

wi = the price of the factors of production, defined as below:

w1 = the price of purchased funds: interest expenses/total deposits and short-term funding.

w2 = the price of labour and physical capital: non-interest expenses/fixed assets.

T = the time trend that captures the influence of technological changes that lead to shifts in the cost function over time.

ε = error term.

From Eq. (3), the marginal costs can be derived as follows:

$$MC_{TAit} = \frac{{\partial TC_{it} }}{{\partial TA_{it} }} = \left( {\beta_{1} + \beta_{2} InTA_{it} + \mathop \sum \limits_{j = 1}^{2} \beta_{2j} Inw_{it}^{j} + \gamma_{4t} T} \right)\frac{{TC_{it} }}{{TA_{it} }}.$$
(4)

Using marginal costs and price, the study calculated Lerner index for each bank and for each year and thus obtained a direct bank-level measure of competition. The index ranges between 0 and 1, with zero reflecting perfect competition and high values corresponding to less competition or high market power.

The study used Panzar–Rosse statistic (H-statistics) as an alternative measure of competition. H-statistics is defined as the elasticity of revenue with respect to the marginal cost of inputs used in the production of banking services (Jeon et al. 2011). H-statistics examines the extent to which a change in input price is translated in the equilibrium revenues earned by a specific bank. It provides a measure referred to as H-statistic which ranges between 0 and 1. H-statistics is based on the responsiveness of banks’ revenue to changes in factor input prices (Schaeck and Cihak 2012). H-statistics is estimated from reduced form of bank revenue equations. It measures the sum of the elasticities of the total revenue of the banks with regards to the banks’ input prices. H-statistics measures competition by assessing the extent to which a change in the factor input prices reflects in equilibrium revenues. The study followed the approach adopted by Fosu (2013) to calculate H-statistics by estimating the following reduced-form revenue equation:

$$In\left( {Pit} \right) = \alpha + \beta 1In\left( {W1,it} \right) + \beta 2In\left( {W2,it} \right) + \beta 3In\left( {W3,it} \right) + \gamma 1\ln \left( {\Upsilon 1,it} \right) + \gamma 2\ln \left( {\Upsilon 2,it} \right) + \gamma 3\ln \left( {\Upsilon 3,it} \right) + \delta D + \varepsilon it,$$
(5)

Pit: is the ratio of gross interest revenue to total assets (proxy for output price of loans)

W1,it: refers to the ratio of interest expense to total deposit as proxy for input price of deposits

W2,it: refers to the ratio of personnel expenses to total assets (proxy for input price of labour)

W3,it: refers to the ratio of other operating and administrative expenses to total fixed assets (proxy for input price of equipment / fixed capital). The model has a number of control variables as captured below

Y1,it: is the ratio of equity to total assets (EQTA) proxy of bank’s leverage

Y2,it: is the ratio of loans to total assets (LTA) account for credit risk exposure

Y3,it: is the total assets (LTA) to control for potential size effects

D: is a vector of year dummies

The subscript i denotes bank i and the subscript t denotes year t. H-statistics is equal to β1 + β2 + β3.

H-statistics is interpreted as follows:

If H-statistic is zero or negative (H ≤ 0), it indicates pure monopoly. This implies that an increase in factor price leads to a fall in revenue. If the value of H-statistic is between zero and one (0 < H < 1), it indicates that banks are in a monopolistic competitive market. Under such a circumstance, an increase in factor price increases average and marginal costs. If H-statistics is equal to 1 (H = 1), it indicates perfect competition where there is free entry and exit. Hence, an increase in factor price leads to a proportional increase in revenue.

The banks’ specific variables employed include firm size (FSIZE), capitalization (CAP), liquidity (LIQ), age (AGE), deposit (DEP), and performance measures (ROA and ROE), while the macroeconomic variables are annual: GDP growth and inflation (INFL) as measured by the consumer price index. Table 3 presents summary definitions of all the variables employed in this study.

Table 3 Summary of variables and measures
Full size table

Empirical specification

To examine the hypothesis regarding the effect of competition on bank risk, the study estimates using a dynamic two-step system generalized method of moments (GMM) as proposed by Blundell and Bond (1998) initially

$$BRT_{{it}} = \alpha _{0} + ~\beta _{1} COMPETITION_{{it}} + ~\beta _{2} FSIZE_{{it}} + \beta _{3} CAP_{{it~}} + ~\beta _{4} LIQ_{{it}} + \beta _{5} AGE_{{it}} + \beta _{6} DEP_{{it}} + \beta _{7} ROA_{{it}} + ~\beta _{8} ROE_{{it}} ~ + ~\beta _{9} GDP_{{it}} ~ + ~\beta _{{10}} INFL_{{it}} + \varepsilon _{{it}},$$
(6)

where \({\text{BRT}}_{\text{it}}\) denotes measures of bank risk-taking for bank i at year t. The BRT indicators are the Z-score, NPLs, LPROV, and CAR. The competition measures are either Lerner index or H-statistics, and \({\varepsilon }_{\text{it}}\) is the error term. Following well-established literature, the study controls for size, capitalization, liquidity, age, and deposit of the banks (Akande et al. 2018; Delis et al. 2016; Berger et al. 2009). The study also controls for the profitability of the banks (ROA and ROE). GDP annual growth (GDP) and inflation (INFL) are the macroeconomics variables employed in the study which are expected to be positively related to the SSA banks’ risk-taking.

Empirical results

Descriptive statistics

Table 4 displays the summary statistics of the variables included in the analysis. Panel A reveals the wide variation of different measures of bank risk-taking. For example, the Z-score ranges from 0.030 to 2.42, with an average (standard deviation) of 0.61 (0.98), which means bank risk-taking values display wide variation which is consistent with the previous studies (e.g., Delis et al. 2016). Panel B provides the summary of competition measures. Lerner index ranges from 0.13 to 0.61 with 0 and 1 representing perfect competition and monopoly, while indices close to 0 or 1 denote monopolistic competition or oligopolistic competition, respectively. The results show that H-statistics figures span from 0.07 to 0.89 with a mean of 0.48. The H-statistics figures are consistent with various literature that has found monopolistic competitive banking systems across SSA region (e.g. Fosu 2013; Fosu et al. 2017). Similarly, the descriptive statistics for bank and country control variables, which are illustrated in Panels C and D, respectively, display wide variation.

Table 4 Descriptive statistics of all variables for all the 2027 bank years
Full size table

Table 5 presents the correlation matrix among different bank risk-taking proxies, competition measures, and the control variables. The correlation analysis reveals that the bank risk-taking proxies (Z-score, NPLs, LPROV, and CAR) positively and significantly correlate with the competition measures (Lerner index and H-statistics). Noticeably, the low correlation coefficients among the variables contained in Table 5 suggest no serious multicollinearity issues.

Table 5 Pearson’s correlation matrices of the variables for the (2027) bank years
Full size table

Multivariate regression analyses

To test whether competition influences bank risk-taking, the study applies the generalized method of moments (GMM) technique. Specifically, Table 6 presents a dynamic two-step system GMM as proposed by Blundell and Bond (1998) of four estimates concerning Z-score, NPLs, LPROV, and CAR. The results meet the various requirements of the regression models, as shown in Table 6. First, the Hansen J-statistics of over-identifying restriction is used to test for the absence of correlation between the instruments and the error term. This tests for the null hypothesis of overall validity of the instruments used. To assess the validity of the instruments used, two additional tests are provided for the existence of first- and second-order serial correlation in the first-differenced residuals (AR1 and AR2) as applied by Coldbeck and Ozkan (2018). The empirical findings of competition measured by Lerner index together with the bank-specific and country control variables on bank risk-taking are provided in Table 6. The coefficients of competition measured by Lerner index on Z-score, NPLs, LPROV, and CAR (0.049, 0.026, 0.054 and 0.236) in Table 6 are statistically positive at 1% (Models 1–3) and 5% (Model 4), respectively. The results provide strong empirical support for H1. The evidence shows that competition measured by Lerner index increases risk-taking in the SSA countries regardless of the bank risk-taking proxies. This result is consistent with the evidence of prior studies that employed Lerner index as competition measure (Leroy and Lucotte 2017; Bushman et al. 2016; Liu and Wilson 2013; Soedarmono et al. 2013; Agoraki et al. 2011; Berger et al. 2009). For instance, Liu and Wilson (2013) examine the effect of competition on risk-taking in a sample of 732 banks in Japan from 2000 to 2009 and observe that Lerner index is positively related with NPLs. The evidence also corroborates the findings of Berger et al. (2009) who conduct similar analysis in an international sample of 8235 banks from 1999 to 2005 and document that Lerner index is positively associated with NPLs. The result is also comparable to the evidence of Agoraki et al. (2011) who investigate the same relationship in 13 European banking systems with 546 banks from 1998 to 2005 and find a positive link between Lerner index and NPLs. In addition, Soedarmono et al. (2013) in a sample of 636 banks in 11 Asian countries from 1994 to 2009 report that Lerner index relates positively with Z-score. Also, Bushman et al. (2016) examine the same link in the US banking industry with a sample of 13,730 banks from 1996 to 2012, and document a positive link between Lerner index and, loan-loss provision and Z-score. Likewise, the findings lend support to the evidence of Leroy and Lucotte (2017) who investigate the same relationship in European banking industry with a sample of 97 banks from 2004 to 2013 and find a positive connection between Lerner index and Z-score.

Table 6 The GMM regression results of the effect of Lerner index on the various bank risk-taking measures
Full size table

Similarly, the results in Table 7 show that competition measured by H-statistics is positively and significantly related to the proxies of bank risk-taking, i.e., Z-score, non-performing loans (NPLs), and loan-loss provision (LPROV) in Models 1, 2 and 3, respectively, thereby offering empirical support for H1. The evidence suggests that an increase in competition increases the risk of loan default while also putting the bank and overall capital of the banks in uncertain circumstances. However, the association between H-statistics and capital adequacy ratio (CAR) in Model 4 of Table 7 is insignificant and does not offer support for H1. Together, the findings suggest that as competition intensifies in the SSA banking system, the propensity of banks to increase their risk-taking also increases. The result of this study is consistent with the findings of prior research that examined the same relationship with H-statistics as competition measure (e.g., Noman et al. 2017; Sarkar and Sensarma 2016). For example, Noman et al. (2017) observe that H-statistics has a positive effect on Z-score in a sample of 180 Asian banks over the period 1990–2014. Also, Sarkar and Sensarma (2016) examine the relationship between competition and risk-taking in the Indian banking industry with a sample of 37 banks from 1999 to 2013, and report a positive link between H-statistics and default risk.

Table 7 The GMM regression results of the effect of H-statistics on the various bank risk-taking measures
Full size table

Overall, the main findings of the study indicate that competition in the SSA banking sector is positively associated with the different measures of risk-taking. The evidence is consistent with competition-fragility view; however, it is in sharp contrast with competition-stability view. Briefly, competition-fragility view articulates that increasing competition in the banking system stimulates bank to engage in excessive risk-taking (Keeley 1990). This is grounded on the argument that increasing competition in the banking system provides several lending avenues to banks which lowers prudent lending. At the same time, an increase in competition decreases the profit of banks, and erodes their charter values and pushing banks to ignore prudent lending, thereby resulting in a deterioration of excessive risk-taking. Within SSA region, the positive competition–NPLs relationship suggests that as competition increases in the deposit market, banks pay higher deposit rate. They face higher repayment burden; hence, the banks charge high loan rate which attract risky investments. Borrowers of the bank therefore assume greater risk, thereby increasing NPLs in the banking system. For example, Tongurai and Vithessonthi (2020) suggest that the probability for borrower’s business ventures to succeed decreases as the interest rates they pay on loans increases. Primarily, banks in the SSA region increase their interest rates due to intense competition in the deposit market. The increase in interest rate reduces earnings on undertaken projects by borrowers (Tongurai and Vithessonthi, 2020). This motivates borrowers to opt for business opportunities that tend to have reduced chances of success. However, these high risky projects tend to have substantial earnings when successful (Tongurai and Vithessonthi, 2020). Subsequently, this shift of borrowers of bank risk-taking behaviour due to rising interest rate may translate into a high NPLs levels in the banking system.

In addition, due to the intense competition in the loan market in the region, the banks tend to have low acceptance criteria for granting loans to attract more demand (Bolt and Tieman 2004). This reduces the quality of the banks’ loan portfolio which gives rise to higher default probabilities in the region. The evidence reaffirms the study of Akande et al. (2018) who show that although deposit insurance which motivates banks to engage in excessive risk-taking is yet to be a popular phenomenon in the region, bank risk-taking in the SSA countries may be linked to moral hazard behaviour. For example, banks in the region tend to act less prudently, because the government and depositors hold responsibility for their actions. The result is consistent with previous investigations that show that competition increases risk-taking in emerging economies (e.g., Noman et al. 2017; Kabir and Worthington 2017; Soedarmono et al. 2013). By contrast, the findings differ from studies that establish a negative relationship between competition and bank risk-taking in developing economies (e.g., Gonzalez et al. 2017; Sarkar and Sensarma 2016; Tabak et al. 2015; Fu et al. 2014; Amidu and Wolf 2013).

Concerning the bank-specific control variables, first, the coefficient of FSIZE which measures the size of the bank has a negative coefficient across all the models. This suggests that larger banks are linked with low risk-taking propensity. This may be in part attributed to the advantage that large banks have in terms of possessing borrower-specific information, superior risk-management capacities stemming from their strong financial position (Toci and Mustapha 2018). Second, other control variables that have statistically significant relationship with risk-taking measures are capitalization, liquidity, and age. For instance, bank age seems to have a positive link with risk-taking measures, suggesting that older banks engage in more risk-taking than younger banks. Liquidity and deposit have positive impact bank risk-taking measures. Thus, it seems that highly liquid banks in the region have a tendency to take more risk in order to grow their market share. The implication is that managers of these banks may choose to deviate from appropriate screening, hence facilitating low-quality borrower’s access to credit. In addition, capitalization has negative effect on the bank risk-taking measures.

Third, the results show that the performance measures (ROA and ROE) have negative and significant impact on the measures of risks. Finally, both GDP and INFL are significantly associated with the risks that the banking sectors of the SSA region face. GDP has an inverse link with Z-score, but is positively related to NPLs and CAR, whiles inflation has positive impact on all the risk measures except NPLs where the link is insignificant.

Robustness checks

The study conducts additional test to check the robustness of the results. Specifically, to control for unobserved firm‐specific heterogeneity, simultaneity, and dynamic endogeneity, the study follows Adu et al. (2022a) and Nguyen et al. (2021) in using a two-stage least-squares (2SLS) approach. Given that the focus of this investigation is on competition and bank risk-taking, this study attempts to find good exogenous instrumental variables (IVs) for these main variables that are correlated with the assumed endogenous variables, but uncorrelated with the error term of the dependent variables (Nguyen et al. 2021). Following the findings of previous studies (Adu et al. 2022a; Nguyen et al. 2021), the study treats the competition measures as endogenous variables. Prior research used 1-year lagged levels of the endogenous variables as primary instruments (e.g., Adu et al. 2022a; Adu 2021). Similarly, the study proposes that lagged Lerner index or H-statistics variables could be appropriate instruments for the analysis. In each of the models, the Durbin score and Wu–Hauman endogeneity tests are used to check the appropriateness of using the 2SLS approach. Specifically, Table 8 provides details about additional checks concerning the impact of competition on bank risk-taking. The study finds similar results in Table 8 and Table 9 as were established in the GMM regression analysis in Tables 6 and 7, respectively. For example, results in Table 8 show that Lerner index has positive and significant impact on all the risk-taking proxies in Models 1–4 (Z-score, non-performing loans, loan-loss provision, and capital adequacy ratio). The results concerning the bank-specific and country control variables reported in Table 8 (Model 1–4) are also comparable with the findings in Table 6 (Model 1–4).

Table 8 The 2SLS regression results of the effect of Lerner index on the various bank risk-taking measures
Full size table
Table 9 The 2SLS regression results of the effect of H-statistics on the various bank risk-taking measures
Full size table

Similarly, the results reported in Table 9 (Model 1–3) reveal that H-statistics has positive effect on Z-score, non-performing loans, and loan-loss provision, respectively. However, the positive impact of H-statistics on capital adequacy ratio (Model 4) is insignificant. The findings of these additional analyses demonstrate that the results do not appear to be driven by any potential endogenous sample selection problems.

Conclusion

This paper provides a cross-country investigation of the impact of competition on bank risk-taking. Specifically, using bank-level data of 220 banks from 16 Sub-Saharan Africa countries over the period 2007–2018, the study constructs competition indicators, risk-taking measures, and bank-level and country-level control variables. The competition measures are Lerner index and H-statistics. The study employs Z-score, non-performing loans (NPLs), loan-loss provision (LPROV), and capital adequacy ratio (CAR) as the dependent variables. After controlling for macroeconomic conditions and bank characteristics, the study finds clear support of a linear relationship between competition and bank risk-taking in the Sub-Saharan Africa banking context.

Specifically, the study detects that competition has a highly significant positive effect on bank risk-taking in the Sub-Saharan Africa countries. The Lerner index has positive and significant impact on the Z-score, non-performing loans, and loan-loss provision which implies that competition is associated with higher bank risk-taking in the Sub-Saharan Africa region. These results are robust to alternative measures and models. This result contributes to the competition-fragility view literature for emerging countries. The main conclusion of the study is that intense competition in the banking system in the Sub-Saharan Africa countries encourages the banks to engage in excessive risk-taking.

The findings highlight various policy implications in the Sub-Saharan Africa economies. First, based on the results, the study concludes that regulators and policymakers in the Sub-Saharan Africa countries need to consider banking competition as very important in the quest for robust banking systems. With the benefits of hindsight, the study contends that the liberalization of the financial sector in the region might potentially have led to intense competition and excessive bank risk-taking. Hence, policymakers should design tailored policies that may curtail excessive competition in the Sub-Saharan African countries. In particular, policymakers may design optimal financial liberalization policies such as mergers and acquisitions to restrict intense competition in the banking system. For instance, regulators may encourage mergers between foreign and domestic banks in the region to improve stability. Second, to prevent excessive risk-taking, regulators and policymakers in the Sub-Saharan Africa countries may impose stricter capital requirements and more enhanced banking supervision to limit bank risk-taking behaviour. Third, the findings of the study lend support to the call by Ali et al. (2022) for researchers to adopt multiple measures of bank competition as a single measure of competition may not adequately capture the impact of competition on bank risk-taking.

The study provides a fertile ground for further research and improvements in the field in the region as highlighted below. First, the study examined the direct impact of competition on risk-taking; however, future studies can expand this study by investigating how capital requirements and bank competition may affect bank risk-taking in the Sub-Saharan African countries. Second, future studies may offer more insights by exploring the joint impact of banking regulation and competition on bank risk-taking in the Sub-Saharan African countries. Finally, given the likely impact of foreign banks on bank risk-taking in the region, future studies may investigate the impact of the entry of foreign banks on risk-taking behaviour in the region.