Determinants of Banks Profitability: Empirical Evidence from Ghana’s Commercial Banking Industry

Over the years, Ghana’s commercial banking industry has been bedeviled with numerous challenges. The unbridled effect of this is the 2018 banking sector megrim which led to the collapse of seven major banks. This pointed out that it is very crucial to identify and mitigate the factors that negatively affect the performance of the banking sector. This paper is used to investigate the effect of banks specific variables (BSVs) and macroeconomic variables (MEVs) on the profitability of commercial banks (NIM, ROE, and ROA) in Ghana using FRED annual data of 25 years. In order to avoid endogeneity problems and aggregation bias, we used the SURE model to run the estimates simultaneously. The result reveals that profit earned by Ghana’s commercial banks is largely influenced by both internal factors such as KA, AQR, LMGT, MEFFI, and Z-Score and fluctuations in the macroeconomic environment (GDP and FOREX). The impact of KA, LMGT, MEFFI, and Z-score is significantly positive whereas AQR (NPLs) is found to have a negative effect on banks profitability. GDP has a significant negative impact on Ghana’s commercial bank's profitability whiles forex induced commercial banks profitability positively, but inflation CPI does not determine the profitability of commercial banks in Ghana. (LMGT), (MEFFI) macroeconomic GDP, FOREX and INFCPI. The dependent variables employed in the model were the bank’s profitability variables. These include return on asset (ROA), return on equity (ROE) and net interest margin (NIM).


Introduction
The banking sectors all over the world act as the life blood of modern trade and economic development and through being a major source of finance to the economy (Murerwa, 2015). The better their financial performance are, the more the shareholders for their investment rewards (Ongore et al., 2013). This, in turn, encourages additional investment and brings about economic growth. It is against this backdrop that the banks performance has been a crucial issue for bank managers, banking regulatory authorities, government, and academic researchers.
The banking sector in Africa, like the rest of the developing world, has experienced major transformation in its operating environment over the last decades. In several African countries, financial sector reforms have been implemented. The commercial banks in Sub-Saharan Africa (SSA) have become more profitable than the rest of the world with an average Return on Assets (ROA) of 2 percent over the last 10 years, significantly higher than bank returns in other parts of the world (Flamini et al., 2009). As one of SSA country, Ghana also embarked the financial sector reforms in the late 1980s as part of ongoing Economic Recovery Program (ERP) (Nkegbe and Ustarz, 2015). Nevertheless, these reforms could not stop seven major banks to collapse in recent years in Ghana. The megrim causing Uni Bank, UT Bank, Capital Bank, Sovereign Bank, Royal Bank, Beige Bank Limited and Construction bank (GH) Limited to cease to operate pointed out that it is very crucial to identify and mitigate the factors that negatively affects the performance of the banking sector in Ghana.
Some researchers have claimed that for Ghana's commercial banks to perform effectively the Bank of Ghana (BoG) should undertake rigorous measures to reduce the monetary policy rate significantly (Soylemez and Ahmed, 2018). It is evident that, a fall in the monetary policy rate will lead to a significant jump in profitability and liquidity of the banks. This is because the banks will have the penchant for lending out more loans to its creditors at a reasonable rate and thus enable these potential borrowers to borrow at a lower interest rate from the commercial banks. This bolsters the performance of businesses in the economy. Some other researchers held contrary view to that recommendation. For instance, according to research conducted by Ndoum (2017) conglomerate has stressed that reduction in the monetary policy rate (MPR) will continue to have marginal effect on commercial banks rate and its performance if the factors specific to the commercial banks rate do not improve. Ndoum (2017) suggested that the key determinants of commercial banks performance have largely been characterized by high non-performing loans (banks asset quality ratio), high operating costs, high cost of funds and high risk of defaults. Ndoum (2017) further indicated that external factors such as inflation, treasury bills and the monetary policy rate (MPR) also contribute to the performance of commercial banks in Ghana.
The determinants of the bank profitability have been investigated in a great deal of studies in the related empirical literature so far. 1 Overall, in the literature, the factors influencing banks profitability are put into two main categories: bank's internal factors (micro-determinants) and external factors (macro-determinants). The banks micro-determinants are generally referred to as the bank's specific factors whereas the macro-determinants are perceived to be the external factors or macroeconomic factors that induce the performance of commercial banks. On the other hand, the internal factors are individual bank characteristics which affect the bank's performance. These factors are basically influenced by the internal decisions of management and board. Some of them are listed as capital adequacy, asset quality ratio, management efficiency, liquidity management, cost efficiency, and ownership identity (Ongore et al., 2013;Wong et al., 2007). On the other hand, the external factors are sector wide or country wide factors which are beyond the control of the banks (Ongore et al., 2013). As external factors, there are resilient macroeconomic environment such as stable inflation and exchange rate as well as sustained increase in the growth of GDP as well as market structure. Among the various approaches, several studies have focused on the structureperformance relationship of banks, the structure-conduct-performance (SCP) hypothesis and the efficient-structure (EFS) hypothesis (Wong et al., 2007).
In this study, we investigate the key factors that affected the profitability of Ghana's commercial banks (35 commercial banks) over the period 1992 to 2017. To this end, we apply SURE model due to its several advantages over the competent models. For instance, the advantages of the model suggest that, while OLS technique can give inconsistent results in case of endogeneity problem, SURE can be used to circumvent this problem by estimating all equations simultaneously (Rachna and Majumdar, 2014). Furthermore, the model avoids aggregation bias and thus ensures a prudent test of equality of regression coefficient vectors applied in the analysis of micro-investment relations (Zellner, 1962), while "the least squares residuals may be used to estimate consistently the elements of covariance matrix of disturbance (Greene, 2002).
A lots of studies such as Ahiabor (2013), Kutsienyo (2011), Nkegbe and Ustarz (2015), Owusu-Antwi et al. (2015), Boadi et al. (2016) and Amo (2015) has been conducted using Ghanaian data to measure the performance of banks profitability. Interestingly, our study is unique and novel in terms of the econometric model applied and the up-to-date nature of our data. The contribution of this study will add to existing literature and make a significant stride to uncover the recent megrims that befuddled Ghana's commercial banking industry.
The rest of the paper is structured as follows: Section 2 provides vivid explanation on the research methodology employed by the researchers and operationalization of the study area. The third section of the paper provides empirical analysis of the data used in the research and the results are presented and discussed in relation to another research. Finally, section 4 provides some concluding remarks and policy recommendation.

Data Set
For this study, we utilized the annual data covering the period between 1992 and 2017, constituting the bank's net interest margin (NIM), bank's return on assets (ROA), bank's return on equity (ROE), asset quality ratio (AQR), capital adequacy ratio (KA), liquidity management (LMGT), management efficiency (MEFFI), bank's z-score, gross domestic product (GDP), inflation (INFCPI) and foreign exchange rate (FOREX) of Ghana. Therefore, the microdeterminants we use in this study are the asset quality ratio (non-performing loans), capital adequacy, liquidity management, management efficiency, banks z-score as, while the macro-determinants are inflation CPI, GDP, and forex. The descriptions and the sources of the variables are reported in Table 1.

Independent
Variables AQR AQR is defined as the commercial bank's asset quality ratio. In Ghana banking industry, it is also known as the bank's non-performing loans (NPL). This measures the ratio of defaulting loans (payments of interest and principal Percent (Worldbank, 2018e) past due by 90 days or more) to total gross loans (total value of loan portfolio). The loan amount recorded as nonperforming includes the gross value of the loan as recorded on the balance sheet, not just the amount that is overdue.

KA
KA represent the capital adequacy of deposit takers. It is a ratio of total regulatory capital to its assets held, weighted according to risk of those assets.
Percent (Worldbank, 2018f) LMGT LMGT represent the commercial bank's liquidity management. This measures commercial bank's ability to meet its financial commitments or obligations. It is a process of effectively managing a bank portfolio mix of assets, liabilities and when applicable off-balance sheet contracts. This process involves two primary financial risks, interest rate and foreign exchange, and directly relates to sound over all liquidity management.

MEFFI
MEFFI defines the management efficiency of commercial banks. Management efficiency is the ratio. between management results (numerator) and management inputs (denominator)In the banking industry, managerial efficiency ensures prudent measures of bank's products such as deposits raised, advances disbursed, and a host of services rendered to depositors, borrowers and others who utilize bank 92 services. Improvement in efficiency will ultimately lead to larger profits and lower costs. The average profit and cost per employee are also taken as indicators to measure the efficiency of employees. The quantum of non-performing asset also plays a major role in deciding the management efficiency of the banks.

Bank's Z-Score
It captures the probability of default of a country's banking system, calculated as a weighted average of the z-scores of a country's individual banks (the weights are based on the individual banks' total assets). Z-score compares a bank's buffers (capitalization and returns) with the volatility of those returns. It is calculated as (ROA+ (equity/assets))/sd (ROA); sd (ROA) is the standard deviation of ROA.
(Calculated from underlying bank-by-bank unconsolidated data from Bankscope) Z-score (Worldbank, 2018g) GDP GDP is the gross domestic product for Ghana. Its measures the sum of gross value added by all domestic producers in the economy plus any product taxes and deduct any subsidies excluded in the value of the products. It is determined without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources.

Methodology: Seemingly Unrelated Regression Equation (SURE) Model
In this study, we employed the seemingly unrelated regression equation (SURE) model to investigate the determinants of bank profitability using data from the Ghanaian economy. SURE model, proposed by Zellner (1962), is used to estimate the parameters of a set of regression equation. Zellner (1962) and Zellner and Huang (1962) set out the seemingly unrelated regression equation (SURE) model as expressed below Here (4) represent U'th equation of an N equation regression on system with a TX 1 vector of observations on the u'th "regressand" variable, a TX I matrix with rank 1 of observations on 1, "explanatory" no stochastic variables, a X 1 vector of regression coefficients and is a TX 1 vector of disturbance terms, each with mean zero.
The system of which (4) is an equation may be expressed as: From equation (4) we generate (3) as where y represents banks profitability in Ghana and X represent both bank specific variables and external factors such as the macroeconomic environment. We further express the model as 0 2E+10 4E+10 6E+10 8E+10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 year GDP INFCPI FOREX Where BPVs represent bank profitability variables such as net interest margin (NIM), Return on Assets (ROA) and Return on Equity (ROE), BSVs denotes banks specific variables such as AQR (Assets Quality Ratio) also known as non-performing loans, KA (Capital Adequacy), LMGT (Liquidity Management), MEFFI (Management Efficiency) and Banks Z-Score. Conversely, the macroeconomic variables are represented by MVs in the model. This includes GDP, Forex, and Inflation CPI (INFCPI). From equation (4) the following equations can be deduce as We then determine the model for each of the bank's profitability indicators as: Alternatively, we rewrite the above models as ∑ Given the variables, the full system of equations is then expressed by: From equation (2)  ] and X represent the block-diagonal matrix on the right-hand side of equation (5). The MT X 1 disturbance vector in equation (5) and (6) are assumed to have the following Variance-Covariance matrix: , where I is the unit matrix of order T X T and ( ) for t=1, 2…, T and μ, μ' =1, 2…M. By transforming equation (5) and (6), we create a single-equation regression model and then employ Aitken's generalized least-squares by premultiplying both sides of (6) by a matrix H which yields E (Huu'H') =H2H'=1. This is given by: ( ∑ ) ∑ Following from (Greene, 2002), if the matrix is unknown, the above model can be estimated by the regression residuals with the consistently estimated elements of ∑ given by j, k 1, ........, where is the least square residuals from equation j.

Empirical Results and Discussions
In this research, in order to investigate the determinants of banks profitability, we applied the lag length selection criteria to determine the number of lags to be used in the model as highlighted in Table 3. We then compute ADF and Phillips-Perron Unit Roots in order to ascertain the stationarity between the variables and thus justify the stability condition of the model as shown in the table 4, 5 and appendix 3. Finally, table 6 exhibit the SURE model computation and empirical analysis. The details of the analysis are highlighted below: Based on AIC and SC we select the most suitable lag of 1. The results of AIC and SC are 14.50732 and 15.09635 respectively indicates the lag order selected by the criterion. They are widely used in VAR modeling. The reason is that the two information criteria are good and reliable and used to choose parameter that minimizes it. The SC selects fewer lag, while any further regressor raises the penalty for the loss of degrees of freedom whiles AIC delivers extremely large model. It is always better to select model with too many lags than too few. This is to make sure that autocorrelation in the remaining VAR model is killed; AIC is therefore chosen as the leading indicator. MacKinnon (1996) approximate one-sided p-value: with Trend and Intercepts at second difference: -4.416345 (1%), -3.622033 (5%), and -3.248592 (10%). As shown in Table 4, the test statics and the one-sided p-values indicates that all the variables were not stationary at level and first difference with trend and intercepts-that is they were not integrated at order zero [I (0)]. This means that there exists unit root among some of the variables i.e., ROA, AQR, LMGT, Z-score, GDP, INFCPI and Forex. In order to use such variable to generate regression coefficient that are unbiased and efficient they must be made stationary. Consequently, the second difference of the NIM, ROA, ROE, AQR, KA, LMGT, MEFFI, Z-score, GDP, INFCPI and Forex were used, and Augmented Dicky-Fuller test was carried out on the variables. The details of the outputs are shown in the Table 4. Above We determine the stationarity of the time series data by running Phillips-Perron tests. At level and 1 st difference with trend and intercepts, no series for PP test were stationary. By determining the second difference of all the series, the result revealed that the time series is stationary; absence of unit root, with the PP test statistic being significant at the 0.01 level at the second difference which is confirmed by the ADF test of revealing significant at 0.99 confident level. With the evidence that, the eight variables are integrated of order one i.e. I (1)they are stationary for both ADF and PP at second difference. Hence it is imperative to run the test for multivariate cointegration using (Johansen, 1988) cointegration test of non-stationary series variables. However, in order to avoid endogeneity problem and aggregation bias we used the SURE model to run the estimates simultaneously. The idea is to check whether the variables influence banks profitability negatively or positively. The unit root test for both ADF and Philip Peron (PP) test were performed to deduce the possibility of stationarity between/amongst the variables. This is indicated in the table 4 and 5. Above NB: ****0.01, ****0.05 and ****0.10 asymptotic critical values for both ADF and PP test at levels at second difference are -4.416345, -3.622033 and -3.248592. NIM which is one of the most important banks profitability in Ghana is positively influenced by commercial bank's internal factors such as Liquidity management (LMGT), Z-score and management efficiency (MEFFI). On the other hand, the net interest margin is also determined by the macroeconomic environment such as Gross Domestic Product (GDP) and foreign exchange rate (FOREX). The impact of GDP to NIM reveals a negative impact on profitability whereas forex contributes positively to the growth of net interest margin (NIM). However, the net interest margin (NIM) is not determined by factors such as capital adequacy (KA), asset quality ratio (AQR) and inflation CPI (INFCPI) since it is not significant at 5% level (marked **).
ROE is influenced by all the internal factors such as KA, AQR, LMGT, MEFFI and z-score and some external factors such as GDP and FOREX determines the ROE among commercial banks in Ghana. However, both AQR and GDP contributes negatively to commercial bank's return on equity.
ROA is also one of commercial banks profitability measures. As can be seen from Table 5, ROA is determined largely by all the internal factors except KA and the macroeconomic variables such as GDP and FOREX does influence the bank's return on assets (ROA). AQR and GDP influence bank's ROA negatively. That is, that inflation CPI (INFCPI) does not determine or contribute to bank's return on assets in Ghana.
The SURE model expounded by Zellner (1962) and Zellner and Huang (1962) shows that the profit earned by Ghana's commercial banks is largely influenced by both internal factors such as the bank's specific variables and fluctuations in the macroeconomic environment. From table 6. Indicates that the key bank's specific variables that determines the bank's profitability were KA, AQR, LMGT, MEFFI and Z-score. However, the AQR shows a negative contribution to the profitability of commercial banks in Ghana. Among the macroeconomic variables that influence the profit earned by commercial were GDP and FOREX, but INFCPI does not in any way determines the profitability of banks in the Ghanaian economy.
St. Err refers to the Windmeijer (1992) robust standard errors, coef = coefficients of the variables, P>|z| = probability values greater than Z-value Superscript ** denotes the acceptance of the null hypothesis that the variable determines banks profitability at less than 5%, significance level meaning that it is statistically significant; [95% Conf. represent the confident level at 95%. Sample included observations: 26

Conclusion and Policy Recommendation
In view of the numerous challenges that befuddled Ghana's commercial banking industry, it is important to critically interrogate the determinants of banks profitability in a lower-middle income country using SURE model. The explanatory variables used in the model constitutes both banks specific variables i.e., capital adequacy (KA), asset quality ratio (AQR), liquidity management (LMGT), management efficiency (MEFFI) and z-score and the macroeconomic variables which encapsulates GDP, FOREX and INFCPI. The dependent variables employed in the model were the bank's profitability variables. These include return on asset (ROA), return on equity (ROE) and net interest margin (NIM).
The empirical findings of the data reveal that both internal factors and external factors contribute immensely to determining bank profitability in Ghana. With regards to the internal factors such as bank specific variables, indicates that KA, LMGT, MEFFI and Z-score contribute positively to determining banks profitability in Ghana. However, the AQR shows a negative contribution to the profitability of commercial banks in Ghana. Meaning that non-performing loans influence the profitability of commercial banks negatively and thus have a significant negative impact on the performance of commercial banks. Among the macroeconomic variables that influence the profit earned by commercial were GDP and FOREX, but INFCPI does not in any way determines the profitability of banks in the Ghanaian economy. However, the analysis reveals that GDP of Ghana contribute significantly to the performance of banks in a negative way and Forex contribute positively to determinants of commercial banks profitability in Ghana. This means that fluctuations of forex or the depreciation of the Ghanaian cedi against the major foreign currencies shackles the progress of banks performance in Ghana. In conclusion, the findings of our studies support the evidence provided by Rachna and Majumdar (2014) and Gunnarsdóttir and Mostepan (2013) The study offers the following recommendation for policymakers to help avert the numerous challenges that negatively affects the performance of commercial banks in Ghana. First, monetary policy regimes aim at ensuring price stability and the stabilization of the local currency by the Bank of Ghana could help deal with the pestilent effects of the fluctuations of the macroeconomic environment.
In a case of asset quality ratio or non-performing loans (NPL), it is imperative for commercial banks to properly sensitize borrowers on the need for timely repayment of their debts and on their rights under the Central Bank of Ghana's Guideline commonly known as the Disclosure and Products Transparency Rules for Credit Products and Services which is in pursuance to Section 7 of the Borrowers and Lenders Act, 2008 (Act 773).Also, frequent publication of the list of commercial banks defaulters and providing incentives for borrowers who pays their debt on time could assist in averting the problems of non-performing loans (NPLs) in Ghana's commercial banking industry.
However, 18 banks constituting 51.43 percent are domestic banks meaning that the ownership rights of the banks are controlled by Ghanaian businessmen and women. The powers and executive rights of those banks are not owned or controlled by foreign banks. Their results reveal that the bank's profitability in Pakistan is explained by size, higher solvency, financial structure, operating cost, labor productivity, market power, and economic growth. They also found an inverted Ushape relationship between banks size and profitability. (Serhat et al., 2018)  The findings indicates that market share of loan is found to be positively related to performance, confirming the relative market power hypothesis and Ghanaian banks pass on their inefficiencies to their customers by raising their lending rates and lowering their deposit rates (Nkegbe and Ustarz, 2015) Ghana Market share of loan is found to be positively related to performance, confirming the relative market power hypothesis and thus determines the banks profit in Ghana. (Nouaili et al., 2015) Bank performance is positively related with capitalization, privatization, and quotation. While, bank size, concentration index and efficiency are negatively related with performance indicators (measured by net interest margin, LIQ, return on assets and return on equity). (Selma et al., 2015)  The results reveals that CAMELS except liquidity and macroeconomic factors affects banks profitability (ROE and ROA) i.e., impact of capital adequacy is significantly positive while Management efficiency and sensitivity to market risk is found to have a significant negative impact on both ROA and ROE. GDP and interest rate has a significant negative impact on both ROA and ROE whereas contrary to our expectation, debt-GDP has a significant positive impact on ROA (Nassreddine et al., 2013)  Bank's Z-score and Stability inefficiency

Appendix 2 Similar Studies in the Review of Literature
The results show that higher insolvency risk/lower bank stability leads to higher profitability of Chinese commercial banks and that higher profitability leads to higher bank fragility. (Mukaila and Mudashiru, 2013) Nigeria

Cointegration and Error Correction Technique
Bank size and cost efficiency did not significantly determine bank profitability. However, credit risk (Loan Loss Provision-Total Assets) and capital adequacy (Equity-Total Assets) was found to be significant drivers which affected bank profitability both in the long run and short run, respectively. Also, while liquidity affected bank profitability in the short run, labour efficiency (Human Capital ROI and Staff Salaries-Total Assets) only affected bank profitability in the long run. But as for the external or macroeconomic variables which determined bank profitability, only Broad Money Supply growth rate was found to be a significant driver both in the long run and in the short run. (Gunnarsdóttir and Mostepan, 2013)  The results suggest that variables that have a significant effect on profitability are leverage, required reserves, inflation, and GDP per capita. Leverage has positive effect