“The interplay of competition, regulation and stability: the case of Sub-Saharan African commercial banks”

ARTICLE INFO Joseph Olorunfemi Akande, Farai Kwenda and Dev Tewari (2019). The interplay of competition, regulation and stability: the case of Sub-Saharan African commercial banks. Banks and Bank Systems, 14(1), 65-80. doi:10.21511/bbs.14(1).2019.07 DOI http://dx.doi.org/10.21511/bbs.14(1).2019.07 RELEASED ON Wednesday, 13 February 2019 RECEIVED ON Friday, 07 September 2018 ACCEPTED ON Wednesday, 05 December 2018


INTRODUCTION
Tied majorly to high commodity prices.
A healthy interplay of competition, regulation and stability is key to managing a bank system that impacts the development of an economy. The economic reality of the Sub-Saharan Africa (SSA) region is that it has a high economic growth rate in terms of GDP 1 (IMF, 2016; World Bank, 2017). However, this high economic growth rate does not translate to development, as larger proportion of the people still live in abject poverty and deprivation (Watkins, 2014). The situation is exacerbated by low financial intermediation the impacts of economic gains from getting to the people. The available banks are inaccessible by low income group because of excessive cost of borrowing and service charges, which are attributed to lack of competition in the financial system, as in most countries, three to five banks account for the bulk of the financial assets of financial systems (Mlachila et al., 2013). Financial inclusion is dismally low with a number of adults to a bank branch overwhelmingly high (Beck & Cull, 2013;Mehrotra & Yetman, 2015). Although competition is good in a bank system because of its potential to lead to efficient banks that could ameliorate some these problems (Casu, Girardone, & Molyneux, 2015), however, if not managed properly, it could result in financial system instability. Fu, Lin, and Molyneux (2014) provided evidence that excessive competition played a significant role in the 2007-2009 world financial crisis despite regulatory efforts at both international and national levels.
Regulation has been long debated in the banking industry and the events leading to the 2007-2009 financial crises have increased attention to bank regulation across the globe (Allen & Carletti, 2013). The SSA banking systems have witnessed remarkable reforms culminating in various forms of regulations. The thrusts of these reforms that took place in different countries at different times were interest rates liberalization, credit control removals, consolidation, deregulations and prudential management, among others, with focus on more capitalized banks. According to Casu, Girardone, and Molyneux (2015) and Llewellyn (1999), the essence of regulating the banking system is to eliminate monopolistic exploitative tendencies while maintaining the stability of the system. Regulation does not only stimulate competition, but also helps in moderating the excessiveness of competitive activities, as the potential instability that could result in excessive competition often constitutes part of the agitations for banking regulation (Angkinand, 2009 (2011) found that the stabilizing power of regulation diminishes when banks have enough market power to increase credit risk and is reversed for banks that possess moderate market power. Moreover, despite the reforms implemented by SSA countries, their banking sectors are highly concentrated (Mlachila et al., 2013) and potentially unstable due to high incidences of non-performing loans, which are on the increase.
Banking competition brings about a stable and efficient banking sector where there is access to finance, low charges and moderate interest rates spread (Chirwa, 2003 . Moreover, high NPLs threaten the stability of the banks in the region. Despite some of the efforts to regulate the sector, these problems have persisted. High costs of banking and lending rates are being identified as factors militating against banking sectors' financial intermediation role. Consequently, service charges are high, financial intermediation is low and high interest rate spreads stifle investment and savings, curtailing the efficient operation of banks in this region, hence, their inability to finance SSA countries' developmental goals. These pose the enormous challenges of the place of regulation in addressing competition in banking without sacrificing stability of the sector to policy makers. According to Vives (2016a), regulation in terms of conduct and structure has the capacity to alleviate the competition-stability trade-off, but not eliminate it. Matutes and Vives (2000) and Vives (2014) agree that capital requirements need to be tougher with more intense competition, but how well to do this has remained unresolved. This study therefore contributes to literature in this area, as it investigates the interrelationship between competition, regulation and stability in the banking system.
Thus, we applied Structural Equation Modelling (SEM) to simultaneously analyze competition, regulation and stability in banks and showed that competition affects stability via efficiency and that regulation affects stability via competition and efficiency. We produced critical theoretical and methodological insights with substantial implications for the conduct of bank regulation policy. Given the key findings here, to the best of our knowledge, this is the first study of this nature. The rest of the paper is organized as follows: the literature review, theoretical and empirical, is presented in section 1. Section 2 explains the method adopted including data source and description of variables used. Section 3 presents the results, the summary and conclusion are presented in the last section.

Competition and stability
Competition is a complex phenomenon in the bank system. The contestable market theory (Abdelkader & Mansouri, 2013) and the competition-stability hypothesis argue that competition will result in stability. The competition-stability hypothesis envisages an efficient bank system capable of orchestrating the stability of the bank system (Schaeck & Čihák, 2014 . Attempting to reconcile this trade-off/dichotomy has in the past occasioned regulatory efforts. Although structural and non-structural models underlying the thinking for a competitive bank system differ in measurements, they converge on the idea that the structure of banks determines their performance (Bresnahan, 1982;Demsetz, 1973;Hicks, 1935;Shepherd, 1983, among others). Based on the market structure theories, a competitive bank system is expected to be efficient at financial intermediation given that they are compelled to innovate and attract customers that are supposedly faced with alternative choices. Hence, a competitive banking environment should produce a stable bank as argued by the authors of the competition-stability view (Boyd et al., 2009; Schaeck & Cih´ak, 2014, among others). They argue that competition is good, since it is consistent with efficiency, which thus enhances the stability of the banking system. The welfare theorem presupposed by these theories is however associated with pure competition whose features impede optimal resources utilization in the industry, resulting in varying levels of inefficiency. Thus, the obvious reality of the banking industry where price is in most banking lines not equal to marginal cost nor is there free entry and exit plus the presence of differentiated products, make pure competition inappropriate. This can be inferred in the traditional competition-fragility theorist's arguments that excessive competition destabilizes the bank system ( According to these authors, high competition in banking causes banks to lose their market powers, culminating in declining profitability. Hence, they become aggressive to invest in riskier portfolios to recoup the financial losses. Therefore, the proponents of this view posit that this risk-taking behavior will erode bank system's stability. Boyd and De Nicolo (2005), considering the assets side of banks' balance sheets, argue that the risk-shifting effects of banks market power in less competitive market are the ultimate cause of instability in banks as their loan portfolio riskiness rises. Martinez-Miera and Repullo (2010) concur with this view, but add that the higher interest rate charged improves the bank's profitability, that is, the margin effect and, as such, presents a U-shaped argument in the competition and stability relationship.

Competition, regulation and stability
These opposing views of competition and stability relationship presuppose regulatory implications. Finance and economic theories seem to have divergent views on the issues of competition, regulation and stability in the bank system without any of them combining these phenomena. In a review of theoretical and empirical literature, Vives (2016b) concludes that competition is not in any way the reason for fragility in the bank system. While he admits that the existence of heightened competitive banking environment could aggravate the situation, instability in banks and systemic failure in general is a subject of certain banking fundamentals. In other words, banks may fail not necessarily because of competition, hence, highlighting the place of regulation. Hakenes and Schnabel (2011) conducted a theoretical review of capital regulation, competition and stability in banks to analyze capital requirements in situations where banks compete on both the assets and liabilities sides of their balance sheets. They concluded that the ambiguous effects of competition on bank risk taking translate to ambiguous effects of capital requirements on financial stability, as banks chose the correlation of their loans' portfolio. The authors further argued that capital can hurt stability because of its influence on competition. In other words, the stabilizing effect of capital regulation tends to be effective in those situations where the charter value effect dominates, and vice versa. Their model suggests that capital regulation may not be suited in all circumstances to prevent excessive risk-taking in banking. On the other hand, Kim and Santomero (1988), in a single-period mean-variance model of the role of capital in risk control, argue that simple capital ratio regulation is ineffective in ameliorating banks' likelihood of insolvency risk, because it ignores the individual banks' different preference structures and allows risky banks to circumvent the restriction via financial leverage and/or business risk. Bolt and Tieman (2004) did a theoretical dynamic modelling of demand for loan to examine the interaction of competition, bank risk-taking and regulation and concluded that increased competition in the banking industry leads to riskier banking behavior. They argue that it is more beneficial for banks to hold more equity than prescribed by regulators, as the more intense the competition, the greater the risk-taking by commercial banks, the higher the failure rates, hence, the charter value falls.
Hellmann, Murdock, and Stiglitz (2000) concluded that if the liability side of bank balance sheets are competitive enough, banks will invest in risky assets, because, as banks compete for insured deposits, they are able to invest in conservative or in volatile assets while raising outside capital. However, banks are closed if capital becomes negative. Though not an efficient means, as capital decreases banks' franchise value, but they argued that it helps to restore incentives to invest in prudent assets by imposing high capital requirements. According to these authors, this explains the superiority of deposit regulation in contrast to capital regulation in excessive competition control.
Modelling competition for deposits with banks utilizing internal capital, Repullo (2004) insists that deposit regulation does not outperform capital regulation. Advocating minimum capital regulation, Allen and Gale (2004), in a general equation model of financial intermediaries and markets with spatial competition and Schumpeterian competition including contagion, concluded that perceived trade-off between competition and stability arising from the allocative efficiency, which is associated with competition, increased the calls for regulation to ensure the coexistence of competition and stability. The authors argue that the most important means to strike the necessary bal-2 Capital adequacy, Asset quality, Management, Earnings, and Liquidity.
ance is to impose minimum capital regulation on banks, as this will reduce the capital available to them; thus, curtail their risk appetite and compete fairly.
Competition and stability have been argued to be the essence of regulations in the banking sector (Casu et al., 2015). More recent literature is now converging on the views that competition is not the cause of fragility in the bank system, but that its presence, especially when in excess, aggravates the risk of instability (Vives, 2016a). It suggests a structural approach to holistically manage the coexistence of these phenomena. Empirical literature as reviewed both within and beyond the SSA region has used a number of methods to test this relationship, but none have achieved a structural analysis where both competition, regulation, and stability are investigated simultaneously. In addition, emphasis on regulation has been largely limited to capital with little or no considerations for liquidity and asset quality regulations as now contained in the expanded Basel Accord, as well as the bank CAMEL 2 . This is the contribution this study seeks to make to literature by not only incorporating the role of liquidity and asset quality regulation, but also structurally investidate the channels among the variables within the bank system, hence, answer why and how the relationship exists. Hence, the ability to capture the variables simultaneously to analyze both the direction of causality and the direct and the indirect relationship using SEM, thus filling the critical theoretical and methodological gap in literature on the conduct of bank regulation policy. This way, the trio of competition, regulation and stability could be better managed in the bank system for optimal performance.

METHODOLOGY
We modelled the relationship between competition, regulation and stability as well as considered the mediation roles of competition and efficiency in the structure as shown in Figure 1, using the SEM.
Given the foregoing literature, Figure 1 hypothesizes the expected relationship between the var-iables. We expect regulation to directly impact competition, efficiency and stability, on the one hand, and an indirect effect on stability through, competition and efficiency. Furthermore, competition and efficiency are also expected to each independently have direct effects on stability, while competition does also indirectly affect stability. These are represented by the paths in Figure 1 such that X 1 , representing regulation variables, is exogenous to the structure, Y 1 , competition, and on another endogenous observed variable that are associated with the two variables. The Γ matrix associated with the two variables defines the structural coefficient of direct effects of exogenous on endogenous observed variables. The indirect effect between two observed variables through a particular mediating variable is then the product of the structural coefficients in the B and/or Γ matrices along the particular path from the exogenous to the endogenous variables. Therefore, the total indirect effect between two observed variables is the sum of all particular indirect effect through all possible mediating variables. The sum of the direct and total indirect effect components between two observed variables is defined as the total effect as contained in Figure 1.
Concerning SEM, it consists of a vast litany of models, from linear regression model to measurement models and simultaneous equations, providing the possibility of capturing the contemporaneous and simultaneous relationship among variables. SEM therefore is a way of estimation, which is not constrained by any kind of model; a multivariate method that permits the system of equation estimation. Moreover, the power also lies in the ability to measure both direct and indirect causal effects among structural variables, hence, the mediation analysis, thereby allowing chains of conditional relationships to be fitted via path analysis. According to Li (2011), SEM has the capability to control for measurement error and is accordingly able to take on the effects of multiple mediators that are applicable to this study, thus, a utility model most effective in testing mediating effects. Furthermore, with SEM model, the direct and indirect effects in the path diagram of competition, regulation and stability can be estimated. Once found to be fit and stable, the strength of SEM lies in determining the path of causality in a structure with the ability to simultaneously deal with multiple endogenous and exogenous variables in addition to the mediating analysis described above. , this study uses the Z-score as a surrogate for stability. The Z-score is a measure of the extent to which banks profit must fall before their equity becomes negative. In essence, it is a measure of the probability of banks' insolvency and therefore encompasses the banks' overall risk, hence, considered as an appropriate stability measure of banks. In the case of efficiency, we used the stochastic frontier analysis (SFA) to generate the efficiency scores of banks using Frontier version 4.1 (Coelli, 1996), a computer program based on stochastic production functions of Coelli (1992, 1995) written to provide maximum likelihood estimates of different types of stochastic frontier production as independently introduced by Aigner, Lovell, and Schmidt (1977) and Meeusen and van Den Broeck (1977). It accounts for truncated normal assumption, including panel data with time varying efficiencies. Hence, applicable to our unbalanced panel model with firm effects having truncated normal random variables distribution assumption that can vary systematically with time (Battese & Coelli, 1992). The necessary data collected for implementing the output-oriented stochastic frontier analysis include banks pre-tax income as used by Chiou and Porter (2015) and asset book value based on J. Barro and R. Barro (1990) as input and output variables. The choice of regulatory variables is contemplated by the main coverage of the Basel Accord and the bank CAMEL on which adequate data could be sourced. Hence, our regulatory capital is the equity capital ratio (ECR), which, according to Casu et al. (2015), constitutes one of the three measures of bank capital regulations. Liquidity (LQTY) regulatory variable used is the ratio of bank liquid assets to depreciation and short-term funds; according to Moyo, Nandwa, Council, Oduor, and Simpasa (2014), by regulation, it dictates the level of liquid assets banks must hold to meet their routine obligations including the funding of their loan assets. Furthermore, the quality of bank assets portfolio is another highly regulated aspect of banking. We proxy asset quality (AQLTY) regulatory variable with loan loss reserve to net loan assets, which serves as health check for banking assets in terms of the proportion of performing loans with implication for bank's performance and bank's stability.

RESULTS
Appendix B presents the summary statistics of the data used; it provides a brief insight into the nature of the data. The main essence of data distribution in Appendix B is to satisfy the thresholds for implementing the SEM model. Overall, the data are normally distributed as confirmed by the Jarque-Bera statistics in Appendix B validating the use of SEM. Once the model is found to be fit, the expectation is a structural banking model where regulation, competition, efficiency and stability simultaneously interplay to provide insight into how best to manage the well-being of the bank system. This suggests that none of the variables can be treated in isolation, as a decision made on one will have a ripple effect on all the other variables in the system. We presented a proposed recursive CRS model in Figure 1 in section 2 for the commercial bank system of the SSA region. The estimated version of this model is shown in Figure 2, which for SEM contains six observed variables, consisting of three observed exogenous variables, ECR, LQTY and AQLTY, as well as three observed endogenous variables, LERNERI, EFF and ZSCORE. Observed exogenous variables are variables with datasets that are determined outside the model and for the sake of the bank system, issues of regulations are largely exogenously handled. These variables have no arrows pointing at them in the model and they are known as independent variables. Observed endogenous variables also known as dependent variables represent variables with datasets that are determined within the system. The pointed arrows show the directions of causality and the estimates on the path represent the standard coefficients estimates generated using STATA 13. Hence, SEM provides an opportunity to model the bank sys-tem as a structural unit. To the best of our knowledge, this is the first application in a study of this nature and a major contribution to literature.
Appendix C is the correlation matrix showing the degree of association among variables. We found that while there is varying degree of weak association between variables, only capital and stability show a very strong positive association, however, the validity of the model results depends on the fitness of the model as presented in Table 1. 5 In terms of model fitness, many goodness of fit models have been proposed in literature as mentioned in subsection 2.1. But a couple of the indi- 5 Model fitness indices presented in Table 1 were achieved by eliminating the paths of LQTY and AQLTY to LERNERI.
cators presented in Table 1 are considered fundamental for any SEM fitness, because they are adjudged to be the most informative (Barbara, 2001;Fan, Thompson, & Wang, 1999;Marsh et al., 1996;Jaccard & Wan, 1996). In terms of the psychometric properties, all the data measurement model fit showed acceptable results for the observed variables examined. The overall model fitness explained by LR chi2 prob of 0.817 indicated that the model is very fit, as null hypothesis requires the prob to be > 0.05. Moreover, the Root Mean Square Error of Approximation (RMSEA) often used as a better approach for testing the fitness of large datasets of SEM model is found to be 0.000 indicating exact fit or absence of misfit. The rule  of thumb is for RMSEA to be < 0.05 for a model to be considered good fit (Brown & Cudeck, 1993; Marais & Andrich, 2007, among others). PCLOSE is the p-value for testing the null hypothesis that the population RMSEA is not > 0.05, which is expected to be > 0.50, and substantiates the exact fit of our model, as it is 1.000 in our study, a further confirmation of model fitness. The Standardized Root Mean Square Residual (SRMSR) provides absolute measure of fit; it is the standardized difference between the predicted correlation and the observed correlation. Because it is an absolute measure, a value of 0 equals perfect fit, but (Hu & Bentler, 1999) suggests a value < 0.08 as generally considered to be a good fit. Our results show SRMSR not to be different from 0. Other tests of fitness used are Comparative Fit Index (CFI) and Tucker Lewis Index (TLI); they are both incremental measures of fits that examine the discrepancy between the data and the model hypothesized. They range from 0 to 1, and specifically values > 0.95 are considered very good fit, and where greater, they are restricted to 1.

Interpretation and discussion of SEM results
Having been satisfied with model fitness, the proposed CERS model estimates are presented in Table 2 to show the probability values. The boxes in Figure 2 represent the interplay of the exogenous, endogenous and/or mediating variables. The arrows linking variables in the structural model depict the direction of causality between variables in the system displaying the standardized coefficients or the standardized regression weights as shown in Table 2. The underlying goodness of fit indices confirms a good fit of the structural model to the data used. All the variables are observed; in other words, they are variables with datasets. Capital (ECR), liquidity (LQTY) and asset quality (AQLTY) are exogenous regulatory variables, while competition (LERNERI) and efficiency (EFF) are endogenous variables that also predict other endogenous variables, stability (ZSCORE).
Furthermore, all the paths allowed to be fitted by the model fit criteria are found to be statistically significant with signs and magnitude shown in Table 2. Specifically, regulatory variables are shown to have a strong and direct influence on the competition, efficiency and stability variables, except that liquidity and asset quality did not cause competition and/or market power. The path of the direct influence includes: capital to competition through efficiency to stability with both liquidity and asset quality to stability through efficiency. Likewise, both the competition and efficiency variables have a strong and direct influence on stability with LERNERI directly causing ZSCORE with EFF also causing ZSCORE, given the path of both competition and efficiency to stability. Figure 2 contains seven mediating paths: ECR → LERNERI → ZSCORE, having competition partially mediating the effects of capital regulation on stability; LERNERI → EFF → ZSCORE, with efficiency partly mediating; and ECR → LERNERI → EFF → ZSCORE, having competition and efficiency partly mediating the effects. Others are: ECR → LERNERI → EFF, with competition partially mediating the effects of capital on efficiency; ECR → EFF → ZSCORE, having efficiency also partially mediating capital and stability relationship. Furthermore, the path LQTY → EFF → ZSCORE shows efficiency to partly mediate liquidity regulation and stability relationship, while AQLTY → EFF → ZSCORE depicts that the effects of asset quality regulation are partially mediated by the efficiency variable. Overall, all effects (total, direct and indirect effects) on the links among the variables in the CERS model are found to be significant. It therefore suggests the presence of partial mediation in the structural model. Competition partially mediates the influence of capital on stability, efficiency partially mediates the effects of competition on stability, and competition-efficiency partially mediates the effects of capital on stability. In terms of causality, capital regulation causes competition, which in turn causes efficiency and then stability of the bank system. This confirms competition-stability views hypothesis in the commercial banks of SSA region and the reference to competition and stability as the reasons for capital regulation.
This study also extends extant literature by establishing direct and mediating effects among other regulatory variables (liquidity and asset quality), efficiency and stability supporting one of our hypotheses in subsection 2.1. While we do not find any causality/path running from liquidity and asset quality to competition, they however both http://dx.doi.org/10.21511/bbs.14(1).2019.07  Matutes and Vives (2000) and Repullo (2004) arguments that capital regulation may not be enough to ensure the ongoing stability of banks, as results demonstrate the explanatory influences of liquidity and assets quality regulations, notwithstanding the dominance of impact capital regulation (in terms of parameter estimates, see Figure 2 and Table 2).
The major conclusion of this study is that the mechanism of regulation, competition and sta-bility in the bank systems operates simultaneously. The implication is that change in policies targeted on any of the components affects the others either instantly or remotely. A further conclusion from the above findings is on the role of competition-efficiency relationship in partly mediating the effects of regulation and stability. As much as regulation directly impacts stability, the influence becomes mediated by competition and efficiency; this again points to the stance in literature that competition may not be the major cause of instability, but that its presence may aggravate it. Moreover, the mediating/transmitting role of efficiency in competition and stability relationship suggests that competition may not in itself cause stability unless efficiency is engendered. The policy implications are enormous for practitioners and regulators alike with further research needed to unbundle how policies should be crafted to address these issues.

CONCLUSION
The essence of this study was to explore the interplay among competition, regulation and stability to enable us to propose a model that could assist policy makers to deal with the issues of stimulating competition without compromising stability in the banking sector. Based on theories, models and extant empirical literature, we fitted a structural equation model to analyze the contemporaneous relationship among competition-efficiency, regulation and stability for the commercial banking sectors of the SSA region. Having established overall goodness of fit of the model, the results show all variables in the structure to be statistically significant. In other words, decisions that affect the exogenous variables (regulation, competition and efficiency) will have simultaneous corresponding effects on the endogenous variables (competition, efficiency and stability) within the banking system. Furthermore, mediating effects became apparent in the results with competition and efficiency partially mediating the impact of capital regulation on banks stability. Hence, we concluded a strong interrelationship among competition, regulation and stability in the banking system.
Obviously, stimulating competition in the banking system without compromising stability constitutes a major puzzle that bank regulators and practitioners face. Therefore, the application of SEM to simultaneously analyze competition, regulation and stability in banks is apt, the result that competition affects stability via efficiency and that regulation affects stability via competition and efficiency, is as well novel, producing critical theoretical and methodological insights with substantial implications for the conduct of bank regulation policy. Even though the degree of influence among variables differs, especially for capital regulation, we recommend that policies should not consider these issues in isolation, but address them holistically, if the bank system is to function at the optimal capacity that could engender the much-anticipated growth and development in the region. Given the enormous challenge in dealing with the balancing act expected to optimize the working of these relationships, we recommend further research work as an offshoot to unbundle the management and implementation process.