Acta Universitatis Danubius. Œconomica, Vol 14, No 3 (2018)
Reexamining Causal Relationship between Dividend Policies and Commercial Bank Performance: Evidence from 30 SubSaharan Africa Countries
Odunayo Magret Olarewaju^{1}
Abstract: This paper aims to test for causality between two dividend policies (dividend payout and dividend reinvestment plans) and return on equity as a measure of financial performance. Dividend policies issues have been continually debated around the world with mixed results, and yet to date, no definite conclusions have been reached. The study used 250 commercial banks from 30 SSA countries over the period between 2006 and 2015 to run longrun causality tests. The results from the block exogeneity Wald test from the panel vector error correction model, and the pairwise Granger causality test shows that there is a unidirectional causality between return on equity and dividend payout ratio. This implies no causality between dividend payout ratio and banks’ return on equity over the study period. Hence, we conclude that the widely adopted model for the payment of dividends in the SSA banking market is a winlose game, as there is no causality between dividend payment and bank performance. As such, we recommend that other dividend policies that can minimize future financing costs, increase bank assets, and improve the future growth prospects of the region be explored.
Keywords: Dividend policy; SubSaharan Africa; Commercial banks; Causality test; Vector Error Correction Model
JEL Classifications: G21; L10; L21
1. Background
While maximising shareholders’ wealth is the main corporate goal of any firm, how to achieve this is up to the individual firm. This goal is achieved through adequate consideration of other stakeholders’ interests, short and longterm financial planning and the implementation of various policies and strategies. (Andriof et al., 2017) A 2014 International Monetary Fund (IMF) summary noted that the ratio of bank assets to GDP is too low in most SSA countries except for South Africa and Mauritius. The summary and the Global Economy report list the ratio of bank assets to GDP as 44.1%, 18.91%, 18.25%, 14.85% and 12.05% in Kenya, Tanzania, Nigeria, Madagascar and Equatorial Guinea, respectively. This is due to the failure of SSA banks to use their earnings as a source of cheap equity that would enable them to operate at full growth potential, promote financial stability and contribute meaningfully to economic growth. (Mlachila et al., 2013b)
Figure 1. Bank assettoGDP ratio in selected SSA Countries in 2014
Source: Author’s calculations based on data sourced from TheGlobaleconomy.com and The International Monetary Fund
From Figure 1, the two countries with the highest bank assets to GDP ratio were Rwanda (53.8%) and Namibia (50.81%), while the others were below 50%. Over the past two decades, the banking sector in SSA has undergone dramatic changes which have led to accelerated economic growth, with commercial banks making the largest contribution. (Beck & Cull, 2013) According to Mlachila et al. (2013c), most countries in subSaharan Africa have an annual growth rate of 5.25%. While some commercial banks are well capitalized, which enabled them to survive the 20072009 global financial crisis, they are considered underdeveloped and are not in a position to sustain future growth prospects in the region. (Mecagni et al., 2015)
The financial system remains the engine of growth for SSA economies and is bankbased. (Moyo et al., 2014) The banking sector makes up more than 70% of this system and accounts for the biggest share of financial assets. (Akande & Kwenda, 2017, Enoch et al., 2015, Allen et al., 2011) However, in comparison with other regions, SSA banks are immature, underdeveloped, highly concentrated and generally inefficient when it comes to financial intermediation, inhibiting their growth. (Allen et al., 2011; Kablan, 2010; Mlachila et al., 2013a) Hindrances to their development include a low asset base, weak creditors’ rights mechanisms, and limited access to financial services, high poverty rates and the small size of national markets in most countries.
The low asset base and its consequences is one of the major constraints to economic growth in the region. (Mlachila et al., 2013a) Should this problem remain unresolved, it will result in instability, regressive performance, a lack of stakeholder confidence, inability to diversify, huge financing costs, and a poor contribution to economic growth which will threaten the banks’ survival. Since the banking sector is the engine of growth for subSaharan Africa, such challenges will hinder the growth of the entire region. There is thus a need to examine what constitute bank assets and how they can be increased to contribute meaningfully to the economic growth of the SSA region. An optimal dividend policy is a tool that can be used by commercial banks to minimise the total agency costs of debt and equity. (Shao et al., 2013) Jensen and Meckling (1976) identify two types of agency costs, the cost of debt arising from conflict between shareholders and debt holders and that of equity stemming from conflict between managers and shareholders. Both types of agency costs affect bank performance and should be minimised in formulating dividend policy. (Shao et al., 2013) A dividend policy that will minimise agency costs and increase bank performance is crucial to achieve corporate goals. (Patra et al., 2012) The agency costs of both debt and equity that emanate from the separation of ownership must be adequately minimised. (Shao et al., 2013)
It remains unclear why SSA banks still pay dividends and what drives their dividend payout policy even though they are characterised by strong concentration, weak creditors’ rights and a low asset base which are impediments to their operational diversification, sustainability and future growth. According to Nnadi et al. (2013), most African firms, including banks, pay annual cash dividends to their shareholders. As shown in the graph below, SSA commercial banks, they consistently pay dividends and the return on equity (ROE) and assets (ROA) which are measures of performance are decreasing. As at 2010, the averaged dividend payout ratio for the selected countries increased dramatically, while the ROA and ROE fell, but picked up in 2011 when the dividend payout ratio (DPOR) started decreasing.


Combined Graph of Dividend and
Performance b) Dividend payout Ratio in SSA banks
c) Return on Asset of SSA countries d) Return on Equity of SSA countries
Figure 2. SSA dividend and performance graphs
Source: Authors’ computation from data collected from BankScope for 20062015. Note that only 30 SSA countries are represented
Ongoing payment of dividends by SSA banks makes no provision to reduce future financing costs which will increase assets, and boost banks’ performance and growth potential. This is because banks’ earnings are important sources of equity that, if reinvested, lead to safe/healthy banking which promotes financial stability and consequently, enhances economic growth. Shao et al. (2013) posit that dividend policy is a crucial firm decision which leads to a conflict of interest between shareholders and creditors (depositors and debtholders). An optimal dividend policy enables firms to manage the divergent interests of creditors and shareholders. Dividend payments reduce shareholders’ fears of expropriation by managers (insiders), while aggravating creditors’ concerns about expropriation by shareholders. (Byrne & O’Connor, 2017) This is because a rational shareholder is more interested in the share price and dividend income than the riskiness of bank operations. (Mehran et al., 2011)
Notwithstanding, studies such as Agyei and MarfoYiadom (2011), Abiola (2014), Ehikioya (2015), and Abdella and Manual (2016) have revealed the correlation between dividend policy (mostly payout policy) and performance both in the banking and nonbanking sector, even though it is not clear which between payout and retention policies will minimize agency problems and improve performance. Granger (1969) opined that it is meaningful and essential to test for causation as opposed to correlation or regression, because correlation/regression is a relationship that does not necessarily imply causation. Causal analysis eliminates the effect of intervention between variables and shows the cause and effect relationship. (Akinlo & Egbetunde, 2010) The absence of studies in this direction in SSA banking and the continuous debate on dividend policy make it imperative for this study to investigate the causal relationship between dividend policies (both retention and dividend payout) and financial performance of banks in SSA countries. Although it has been established that there is a feedback relationship between dividend policy and bank performance (Hamid, Yaqub & Awan, 2016), the form this policy should take to foster effective performance is still a gap that this study intends to bridge. Therefore, this study weighs the two common dividend policies in banking sector and establishes the causal relationship between each of them and bank performance in the SSA region. Since the focus of this objective is to test causality, block exogeneity Wald and pairwise Granger causality tests were conducted. This is different from the test of feedback relationship that has been done in previous research.
The paper is structured into 6 sections for logical presentation. The next section contains the literature review, which is followed by the theoretical underpinning in section three. The methodology is captured in section four, model estimation in section five, and conclusion in section six.
2. Literature Review
2.1. Empirical Review
Few studies have empirically examined the relationship between dividend policy and performance in the banking sector. Hamid et al. (2016) recent study investigated Pakistan banks’ dividend policy. The results of the FEM showed that tax and financial slack (retained earnings) had no significant effect on bank performance. It is pertinent to examine if retained earnings impact performance despite its insignificant effect. Waseem et al. (2011) examined dividend policy and performance stability among 17 listed commercial banks in Jordan for the period 20002006 using pooled EGLS (crosssectional random effect). The result showed that the cash dividend policy is unstable in Jordanian banks and hence has a negative effect on their performance. Agyei and MarfoYiadom (2011) examined 16 Ghanaian commercial banks for the period 1999 to 2003. Using the fixed and random effect model estimation, they found that, dividend paying banks enhanced their performance. This requires further enquiry because the study was silent on banks that retain earnings to enhance growth. There is a need to weigh both policies to determine which tends to promote bank performance. Evidence from Bangladesh for the period 2008 to 2010 using regression analysis showed that variations in dividend policy did not explain variations in the returns on shares among the listed commercial banks. (Zaman, 2011)
The review of the literature shows that, the majority of studies focused on nonbanking sectors. Ouma (2012) examined Kenyan listed firms for the period 2002 to 2010 using regression analysis. The study found a strong and positive relationship between dividend payouts and firm performance. However, it did not consider other dividend policies. Ajanthan (2013) analysed the effect of dividend payouts on the performance of hotels and restaurants in Sri Lanka using multiple regression estimation. The study found a strong and positive relationship between dividend payouts and firm performance. Priya and Nimalathasan (2013) employed regression analysis to investigate hotels and restaurants in the same country for the period 2008 to 2012. The study found a significant positive relationship between dividend policy and firm performance. However, it did not specify the kind of dividend policy that influenced performance. Evidence from Nigerian manufacturing firms presented by Uwuigbe et al. (2012) and Ehikioya (2015) showed that dividend payouts significantly and positively impacts firm performance. However, both studies neglect retention policy and this requires further clarification as payout is not the only dividend policy. There are always some shareholders that prefer to retain profit to grow the firm and avoid external borrowing to finance viable investment projects. Furthermore, firms with financial constraints pay significantly lower dividends than those that are financially buoyant. (Obembe et al., 2014) Evidences from past research have shown conflicting findings on which dividend policy (payout or retention) should be adopted in firms such that the managers will maximize wealth and not profit even though majority believe banks should pay out. There is a discrepancy in the various theories on the choice of dividend policy and empirical findings have shown that not all firms that pay out have prospects of positive future performance as a result. Studies have shown that dividend policy has a great effect on bank performance, irrespective of the particular policy adopted. However, the policy that will cause performance has not been identified because correlation does not mean causality.
2.2. Theoretical Underpinning
Life Cycle Theory of Dividend
The life cycle theory of the firm was propounded by Mueller (1972) and the life cycle theory of dividends was developed by Bulan and Subramanian (2009) and Thanatawee (2011). The major argument of this theory is that the stage a firm has reached in its life cycle determines its optimal dividend policy. The life cycle runs from inception to maturity with many circumstances arising along the way, including a declining growth rate, shrinking investment opportunities, and the decreased cost of raising capital externally. The optimal dividend policy involves a tradeoff between the costs and benefits associated with raising capital for new projects taking lifecyclerelated factors into account. Dividend policies change over the life cycle of a firm, and, surprisingly, as a firm matures, its ability to explore profitable investment opportunities is overtaken by its ability to generate cash.
The relationship between a firm’s life cycle and dividend policy is shown schematically below:
Stage of Growth Dividend Policy
Introduction Undecided
Rapid Growth Higher retention policy
Maturity Stage Increased payout policy
Declined Growth Generous payout policy
3. Methodology
3.1. Research Design and Data Source
Following, to name but a few, Díez Esteban and López de Foronda Pérez (2001), Flamini et al. (2009), Nnadi et al. (2013), Francis (2013) and Akande and Kwenda (2017) regional studies on African or SSA banks that used unbalanced panel datasets from several commercial banks, this study is based on unbalanced panel of 250 commercial banks from 30 SSA countries. As noted in the previous chapter, all these countries have similar economic and banking features such that their banking markets are oligopolistic in nature. Panel data was used to cater for the heterogeneity problem that individual bank characteristics might cause. (Hsiao, 2014) Not all the data required to capture the variables of interest were available for all the SSA countries for the study period; hence, it was unbalanced. The unbalanced panel data analysis approach was used rather than a balanced panel because we are less interested in goodness of fit and more concerned with understanding the explanatory and illuminating powers of the specific variables, using the available data. The annual data were collected from BankScope database by Fitch/IBCA Bureau Van Dijk covering the period 2006 to 2015. The SSA countries considered in this study exclude those regarded as outliers such as South Africa and Mauritius due their highly competitive and sophisticated banking systems. (Beck & Cull, 2013) Countries such as such as Democratic Republic of Congo, Comoros, GuineaBissau, Sao Tome and Principle and that lacked data due to the effects of war were also excluded. (Akande & Kwenda, 2017; Flamini et al., 2009) The countries selected are bankbased economies in which commercial banking holds more than 70% of financial system assets on average. They have similar economic and banking characteristics such as weak creditors’ rights, underdeveloped infrastructure, high inflation and poverty rates, external shocks, high concentration, a shallow financial system and nonadherence to global regulatory requirements. (Akande & Kwenda, 2017; Allen et al., 2014; Flamini et al., 2009)
3.2. Model Specification
_{Theories such as the birdinthehand theory, the signaling hypothesis, as well as the empirical findings of }_{Agyei and MarfoYiadom (2011)}_{; }_{Zakaria, Muhammad, and Zulkifli (2012)}_{; and }_{Hamid et al. (2016)}_{ have averred the feedback relationship between dividend policy and performance. }
_{Therefore,}


_{(}i) 
_{That is, }
_{Performance = }_{f }_{(dividend policies)}
_{As recently noted by }_{Hamid et al. (2016)}_{, though dividend policies are unstable in the banking sector, there are two commonly adopted policies: dividend payout policy (in form of cash) and dividend reinvestment plan (DRIP). Note that DRIP is otherwise called retention policy. }
_{Hence, }
_{Performance = f (dividend payout ratio, retention ratio)}
_{To avoid the omission of a germane variable which can lead to a simultaneity bias, the capital adequacy ratio is included as a control variable.}


_{(}ii) 


_{(}iii) 
All variables are in their natural form.
_{ }are the estimated parameters of the respective explanatory variables which show the percentage change in financial performance caused by percentage change in the explanatory variables;
_{ }is the intercept/constant term.
ROE = DPOR= , RERA= and CAR= .
3.3. Estimating Technique
Specifically, Pairwise Granger Causality and Granger causality tests from the PanelVector Error Correction Block Exogeneity Wald test were used to establish both the short and long run unidirectional, bidirectional causality or nocausal relationships between the pairs of variables.


_{(}iv) 


_{(}v) 


(vi) 


_{(}vii) 
Where:
_{ }_{are respective constants. } _{ are respective estimated coefficients }∆ denotes the first difference operator. represents the oneyear lagged Error Correction Term. It is the cointegrating vector that acts as the speed of adjustment for the long run association between the variables. are mutually uncorrelated stochastic (white noise) error terms with finite covariance matrix and zero mean value. is the time period that ranges from , is the crosssection (banks) that ranges from _{ }and lastly, k is the number of lags while is the optimal lag length selected by using the Sequential modified LR test statistic, Final Prediction Error, Akaike Information Criterion (AIC), Schwarz Bayesian criterion (SBC), and HannanQuinn information criterion. To conduct a multivariate causality test, , which is the measure of Capital Adequacy Ratio, was included to avoid the omission of germane variables that can cause simultaneity bias and thereby lead to a bogus relationship between the variables. (Gujarati & Porter, 2003) For any commercial bank to adopt a policy, it must be adequately capitalized to justify the continuity of banking activities and hence, their persistent future growth. This follows from the fact that banks avert risk by maintaining a high degree of capitalization.
4. Model Estimation and Interpretation of Findings
4.1. Preliminary Analysis
4.1.1. Summary Statistics
This section shows the description of the characteristics of the variables ranging from the mean, median, maximum, minimum, standard deviation, skewness, kurtosis, and probability of the distribution.
Table 1. Descriptive Analysis

ROE 
RERA 
CAR 
DPOR 
Mean 
0.198431 
0.544922 
0.135358 
0.459001 
Median 
0.222880 
0.567239 
0.112732 
0.437637 
Maximum 
13.88820 
1.000000 
1.073452 
2.670570 
Minimum 
31.53604 
1.670570 
2.067475 
0.000000 
Std. Dev. 
0.859533 
0.241318 
0.133588 
0.237609 
Skewness 
21.07763 
0.849753 
0.005022 
0.866069 
Kurtosis 
827.0631 
7.155175 
58.96471 
7.421875 
Probability 
0.000000 
0.000000 
0.000000 
0.000000 
Source: Author’s estimation, 2018
Table 1 shows the descriptive analysis results of all the activities regarding the causal relationship of dividend policy and bank performance for the period between 2006 and 2015. return on equity (ROE) measured the performance of the banking industry while dividend payout ratio (DPOR), retention ratio (RERA), and capital adequacy ratio (CAR) were used to capture the dividend policy. The result revealed that the average ROE, DPOR, RERA, and CAR are 0.20, 0.46, 0.54, and 0.14, respectively. This result implies that the average performance of the banking industry as determined by the return on equity is small and not encouraging. The maximum & the minimum values for rate of ROE, DPOR, RERA, and CAR are: 13.89 & 31.54, 2.67 & 0.00, 0.98 & 8.31, 1.00 & 1.67, and 1.07 & 2.07, respectively. The standard deviation values of 0.86, 0.24, 0.24, and 0.13 revealed the rate at which the ROE, DPOR, RERA, and CAR, respectively, deviated from their respective average or expected values. Also, it was discovered that CAR and DPOR, which are 0.01 and 0.87 respectively, are positively skewed because their distribution has a long tail to the right. On the other hand, ROE and RERA which are 21.08, and 0.85 respectively, are negatively skewed because their distribution has long tail to the left. However, the kurtosis of the financial variables showed that all the variables under consideration are leptokurtic in nature because the kurtosis coefficient indexes are all positive. The JarqueBera and probability values revealed ROE, DPOR, RERA, and CAR are not normally distributed, but statistically significant when examining the impact of dividend policy on the performance of SSA banking industry.
4.1.2. Panel Unit Root Test
To run analysis on secondary data, it is necessary to run stationarity tests in order to know the nature of data and avoid spurious estimations. As noted by Akinlo and Egbetunde (2010), none of the unit root tests is free from size and power properties shortcomings. Thus, to be affirmative about the order of integration, several unit root tests are conducted in this study as shown in table 2 below.
Table 2. Panel unit root test at level for the variables
Variables 
Levin, Lin & Chu t* Statistic 
Prob 
ADF Statistic 
Prob 
PP Statistic 
Prob 
ROE 
59.6740 
0.0000 
18.4207 
0.0001 
18.4207 
0.0001 
RERA 
2.88111 
0.0020 
136.346 
0.0000 
147.042 
0.0000 
CAR 
5.64249 
0.0000 
195.317 
0.0000 
212.920 
0.0000 
DPOR 
8.75976 
0.0000 
188.866 
0.0000 
163.428 
0.0000 
Source: Author’s estimation, 2018
The panel unit root test presented in Table 2 above shows that all the variables were stationary. The ROE, dividend policy ratio (DPOR), retention ratio (RERA), and capital adequacy ratio (CAR) were all stationary at order one for both crosssection and individual level during the period under investigation. This is evident as the probability of Levin, Lin and Chur tstatistic values: 0.000, 0.002, 0.000 and 0.000; the Augmented Dickey Fuller (ADF) teststatistic and Philip Perron statistic values: 0.000, 0.000, 0.000, 0.000 and 0.000 for each of the variables was less than the probability of the error margin 0.05 allowed for the estimate in this study. This result implies that there is a shortrun equilibrium relationship among the variables under investigation. The shortrun stability of these variables revealed by the panel unit root test led to the estimation of cointegration to determine the longrun equilibrium relationship or stability of the linear combination of the variables in the longrun.
4.1.3. Panel Cointegration Test
Since all our variables at nonstationary at order one (1), a cointegration test was used to establish the longrun relationship between the variables using the Johansen methodology. Johansen’s approach derives two likelihood estimators for the cointegrating rank: a trace test and a maximum Eigen value test. The cointegrating rank was formally tested using the trace and the maximum eigen value statistic.
Table 3. Cointegration Rank Test using Trace Statistic
Eigen value 
Trace Statistic 
5% Critical Value 
Prob 
Hypothesized No. of CE(s) 
0.220906 
1933.973 
69.81889 
1.0000 
None * 
0.165168 
935.2810 
29.79707 
0.0001 
At most 1 * 
0.124231 
527.2958 
15.49471 
0.0001 
At most 2 * 
0.095763 
227.5006 
3.841466 
0.0000 
At most 3 * 
Source: Author’s estimation, 2018
Note that * represents the significance of the test statistics
Table 4. Cointegration Rank Test using Maximum Eigen Value Statistic
Eigen value 
Maximum Eigen Value Statistic 
5% Critical Value 
Prob 
Hypothesized No. of CE(s) 
0.220906 
564.1485 
33.87687 
0.0001 
None * 
0.165168 
407.9852 
21.13162 
0.0001 
At most 1 * 
0.124231 
299.7953 
14.26460 
0.0001 
At most 2 * 
0.095763 
227.5006 
3.841466 
0.0000 
At most 3 * 
Source: Author’s estimation, 2018
Note that * represents the significance of the test statistics
These test statistics indicate three cointegrating vectors at 5 percent level of significance as presented in Tables 3 and 4 above. This implies that a longrun equilibrium relationship exists among the variables under study. Thus, the stability of the dividend policy captured by the DPOR, RERA, and CAR has affected SSA banks’ performance measured by ROE in both the short and longrun. The above tables also show, that the maximumeigen value test indicates three normalized cointegrating equation(s) at 5 percent significant level. The details of these three normalized cointegrating equations and their adjustment coefficients are presented in Table 5 below.
Table 5. Cointegration Equations
ROE 
RERA 
CAR 
DPOR 
1.000000 
0.000000 
0.000000 
0.068027 



(0.03589) 
0.000000 
1.000000 
0.000000 
0.377640 



(0.01003) 
0.000000 
0.000000 
1.000000 
0.182568 



(0.01407) 
Adjustment coefficients (standard error in parentheses) 

D(ROE) 
0.912529 
0.047255 
0.063485 

(0.03621) 
(0.17490) 
(0.08983) 
D(RERA) 
0.005543 
0.369680 
0.205679 

(0.00864) 
(0.04174) 
(0.02144) 
D(CAR) 
0.007637 
0.050936 
0.053140 

(0.00428) 
(0.02069) 
(0.01063) 
D(DPOR) 
0.008845 
0.200160 
0.352320 

(0.00835) 
(0.04032) 
(0.02071) 
Source: Author’s estimation, 2018
Figure 3. Graphical Representation of Cointegrating Equations
Source: Author’s estimation, 2018
Table 5 and Figure 3 present the normalized cointegrating equation(s) coefficients with their standard error in parentheses. The normalized cointegrating coefficients only load on the dividend payout ratio with both positive and negative coefficients. The coefficients of the dividend payout ratio, 0.068, 0.378 and 0.183 are statistically significant based on the standard error test. This implies that banking performance as shown by the cointegrating equations can be determined by futurestate and the stability of ROE, RERA, and CAR while the DPOR mainly determines the current level of banking performance and its movement in the right direction to bring the system back to equilibrium. The cointegration adjusted coefficients measure the longrun equilibrium or stability of banking performance. The ROE value of 0.913 in the first cointegrating equation reveals the performance level of the selected SSA banks which is not encouraging and calls for improvement. The adjustment coefficients values of 0.006 and 0.009 from cointegrating equation one reveal the negative impact of the RERA and DPOR, both dividend policies, on banking performance. However, the capital adequacy ratio value of 0.007 contributes positively to SSA banking performance. In the second cointegrating equation, the performance of the banking industry improved, as the performance level stood at 0.047. This was hampered in the third equation as a result of the negative impact of the DPOR in the first and second cointegrating equation which limits the performance of the banking industry by 0.009 and 0.200, respectively. The negative impact of the retention ratio in the second cointegrating equation also hampered SSA bank performance during the period under study. This implies that the more attention that is devoted to formulating effective dividend policy, the better will be the performance of the SSA banking industry.
4.1.4. Optimal Lag Selection
Table 6. VAR Lag Order Selection Criteria
Lag 
LogL 
LR 
FPE 
AIC 
SC 
HQ 
0 
2654.561 
NA 
9.20e06 
2.593428 
2.607144 
2.598458 
1 
1527.413 
2247.701 
3.14e06 
1.518686 
1.600979 
1.548864 
2 
1437.331 
179.1966 
2.95e06 
1.455223 
1.606093* 
1.510549* 
3 
1403.290 
67.55168 
2.92e06* 
1.446407* 
1.665854 
1.526880 
4 
1385.211 
35.78683 
2.94e06 
1.453156 
1.741181 
1.558778 
5 
1362.029 
45.77771 
2.95e06 
1.454928 
1.811530 
1.585698 
6 
1335.590 
52.07799 
2.94e06 
1.453525 
1.878705 
1.609443 
7 
1321.220 
28.23597 
2.97e06 
1.463891 
1.957647 
1.644957 
8 
1295.670 
50.07920* 
2.97e06 
1.463354 
2.025688 
1.669568 
Source: Author’s estimation, 2018
Table 6 shows the result of the vector autoregressive lag length to choose the optimal lag for this study. The result shows a lag order of three (3) using the Akaike information criterion with a value of 1.446 while the lag order of two (2) using the Schwarz information criterion and HannanQuinn information criterion has values of 1.606 and 1.511, respectively. All these information criteria were statistically significant at 5 percent level. Based on this evidence, lag order two (2) which was the smallest minimum lag order revealed by Schwarz information criterion and HannanQuinn information criterion was selected for this study.
4.2. Panel Vector Error Correction Model Estimation
A Vector Error Correction Model (VECM) is a restricted Vector Auto Regression (VAR) used for nonstationary cointegrated series. Therefore, since our series is cointegrated, a VECM with four (4) simultaneous equations is estimated in this study to evaluate the shortrun properties of the longrun relationships among ROE, DPOR, RERA and CAR. Optimal lag two (2) of SIC was chosen for this estimation as noted by Hyndman and Athanasopoulos (2014) that SIC is the most suitable for VECM as AIC tends to choose larger number of lags that make the estimate insignificant.
Table 7. Vector Error Correction Estimates
Cointegrating Eq: 
CointEq1 
CointEq2 
CointEq3 

ROE(1) 
1.000000 
0.000000 
0.000000 

RERA(1) 
0.000000 
1.000000 
0.000000 

CAR(1) 
0.000000 
0.000000 
1.000000 

DPOR(1) 
0.068027 
0.377640 
0.182568 


(0.03592) 
(0.01003) 
(0.01408) 

C1 
0.131002 
0.183238 
0.038462 

Error Correction: 
D(ROE) 
D(RERA) 
D(CAR) 
D(DPOR) 
CointEq1 
0.912529 
0.005543 
0.007637 
0.008815 

(0.03624) 
(0.00865) 
(0.00429) 
(0.02657) 
CointEq2 
0.047255 
0.369680 
0.050936 
0.504038 

(0.17502) 
(0.04177) 
(0.02070) 
(0.12831) 
CointEq3 
0.108874 
0.199250 
0.270764 
0.512624 

(0.14556) 
(0.03474) 
(0.01722) 
(0.03356) 
CointEq4 
0.063485 
0.205679 
0.053140 
0.352320 

(0.08989) 
(0.02145) 
(0.01063) 
(0.02072) 
D(ROE(1)) 
0.066835 
1.05E05 
0.000862 
0.004100 

(0.03043) 
(0.00726) 
(0.00360) 
(0.00701) 
D(ROE(2)) 
0.034568 
0.002248 
0.000578 
0.000651 

(0.02241) 
(0.00535) 
(0.00265) 
(0.00517) 
D(RERA(1)) 
0.006249 
0.296242 
0.038291 
0.163172 

(0.17328) 
(0.04135) 
(0.02050) 
(0.03995) 
D(RERA(2)) 
0.162397 
0.097199 
0.025696 
0.085151 

(0.14162) 
(0.03379) 
(0.01675) 
(0.03265) 
D(CAR(1)) 
0.014263 
0.156189 
0.100861 
0.119614 

(0.18496) 
(0.04414) 
(0.02188) 
(0.04264) 
D(CAR(2)) 
0.043268 
0.042329 
0.117200 
0.013583 

(0.17573) 
(0.04194) 
(0.02079) 
(0.04052) 
D(DPOR(1)) 
0.012641 
0.012609 
0.003116 
0.001286 

(0.05576) 
(0.01331) 
(0.00659) 
(0.01285) 
D(DPOR(2)) 
0.005704 
0.003621 
0.002557 
0.013647 

(0.04561) 
(0.01088) 
(0.00539) 
(0.01051) 
C2 
8.81E05 
0.000140 
0.000235 
0.000130 

(0.01882) 
(0.00449) 
(0.00223) 
(0.00434) 
Rsquared 
0.493118 
0.267416 
0.190113 
0.288941 
Adj. Rsquared 
0.489957 
0.262848 
0.185063 
0.284507 
Sum sq. resids 
1797.073 
102.3385 
25.14159 
95.52210 
S.E. equation 
0.894694 
0.213507 
0.105825 
0.206274 
Fstatistic 
156.0030 
58.53543 
37.64233 
65.16166 
Log likelihood 
2947.799 
290.3617 
1876.613 
368.2505 
Akaike AIC 
2.621946 
0.243683 
1.647445 
0.312611 
Schwarz SC 
2.659931 
0.205698 
1.609460 
0.274626 
Mean dependent 
0.000383 
0.000275 
0.000289 
0.000223 
S.D. dependent 
1.252770 
0.248675 
0.117227 
0.243860 
Source: Author’s estimation, 2018
The presence of cointegration between variables suggests a longterm relationship among the variables under consideration. The VCEM can then be applied. The vector error correction estimate with standard error in parenthesis for the long run relationship between dividend policy and banking performance for three cointegrating equations is presented in Table 7 above. The C1 in the cointegrating equation are correctly signed, revealing that it will take 13 and 18 percent, respectively for the maladjustment in the cointegrating equation 1, and 2 to attain or adjust to the long run equilibrium or stability. In examining the impact of the error correction of the dividend policy on banking performance, it was found from the fitted VECM that ROE at lag one and two, RERA at lag one and two, and CAR at lag one have an inverse relationship with the banks’ ROE (performance). Thus, ROE at lag one and two, RERA at lag one and two, and CAR at lag one will worsen the banks’ ROE ratio (performance measure) by 6.68, 3.46, 0.62, 16.24, and 1.43 percent, respectively. This implies that particular circumstances in the countries examined during the study period impacted the influence of the dividend retention policy on bank performance because all things being equal, this should not be a negative effect. This finding is contrary to Uwuigbe et al. (2012) and Agyei and MarfoYiadom (2011) studies that found that either payout or retention dividend policy has a positive relationship with performance. However, CAR at lag two, and DPOR at lags one and two have a direct relationship with banking performance. This is in tandem with the findings of Brighi and Venturelli (2014) and Odunga et al. (2013). The results further reveal that CAR at lag two and DPOR at lags one and two improved the performance of banking industry by 4.33, 1.26 and 0.57 percent, respectively. This positive effect reveals the signalling effect of dividends during this period such that it had a direct effect on bank performance as revealed by Ehikioya (2015). The positive relationship between capital adequacy and performance implies the significance of capital adequacy in formulating dividend policy. As noted by Nnadi et al. (2013), a bank must be adequately capitalised before making dividend decisions. The positive relationship further depicts the nexus between the capital ratio and funding costs. A bank with a high capital ratio incurs lower funding costs because of reduced bankruptcy costs. (Brighi & Venturelli, 2013) The C2 estimate of 8.81E05 reveals that the banking industry’s performance could have been enhanced and improved through dividend policy during the period under investigation without serious risk. The significance of the VECM was examined using the Rsquare statistic and it was revealed that 49 percent of the variation in the error associated with the performance of SSA banks can be explained by the dividend policy captured by the RERA, CAR and DPOR. The F statistic value of 156.00 ˃ F_{0.05 }(3, 1714) = 3.00 shows that the fitted VECM was statistically significant and hence adequate and reliable in determining the causal relationship between the dividend policy and banking performance (ROE) in SSA.
4.2.1. PVECM Stability Check
Figure 4. Diagrammatic Representation of VECM Stability Condition Check
Source: Author’s estimation, 2018
The results in Figure 4 show the VECM stability condition for the relationship among ROE, RERA, CAR and DPOR. However, the figure shows that, all the roots or the eigen values were within the unit circle. This implies that the VECM satisfied the stability condition. It can hence be used for policy formulation and implementation.
4.3. Granger Causality Test
According to Fisher (1993), economic theory guarantees causality in at least one direction of any cointegrated series. Therefore, there is need to test for causality in this study. Table 8 shows the block exogeniety Wald test following the study conducted by Gul and Ekinci (2006) where it was posited that causality can be established using probability and chisquare statistics under the null hypothesis of no causality.
Table 8. VEC Granger Causality/Block Exogeneity Wald Tests
Dependent variable: D(ROE) 

Excluded 
Chisq 
D.f. 
Prob. 
D(RERA) 
1.930692 
2 
0.3809 
D(CAR) 
0.087883 
2 
0.9570 
D(DPOR) 
1.756823 
2 
0.4154 
All 
6.394357 
6 
0.6031 
Dependent variable: D(RERA) 

D(ROE) 
0.360031 
2 
0.8353 
D(CAR) 
12.54067 
2 
0.0019*** 
D(DPOR) 
17.95587 
2 
0.0001*** 
All 
37.31987 
6 
0.0000*** 
Dependent variable: D(CAR) 

D(ROE) 
0.061813 
2 
0.9696 
D(RERA) 
3.784374 
2 
0.1507 
D(DPOR) 
19.16546 
2 
0.0001*** 
All 
44.22851 
6 
0.0000*** 
Dependent variable: D(DPOR) 

D(ROE) 
0.947218 
2 
0.6228 
D(RERA) 
16.75244 
2 
0.0002*** 
D(CAR) 
8.226999 
2 
0.0164** 
All 
68.30802 
6 
0.0000*** 
Source: Author’s estimation, 2018
Note that *** and ** represent rejection of Ho at 1% and 5% respectively
The result of vector error correction Granger causality among the financial variables under consideration are presented in Table 8 above to show the direction of causal relations between each pair of the financial variables such as return on equity, retention ratio, capital adequacy ratio and dividend payout ratio. There is a unidirectional causality of error between the capital adequacy ratio and retention ratio; dividend payout ratio and retention ratio; and dividend payout ratio and capital adequacy ratio. It was also found that there was bidirectional causality of error between the retention ratio and dividend payout ratio; and capital adequacy ratio and dividend payout ratio. This is evident in the estimated probability of Chisquare statistic values of 0.002, 0.000; 0.016, and 0.001˂ 0.05. Thus, error as a result of the capital adequacy ratio and dividend payout ratio Granger causes error that arises as a result of the retention ratio; and errors as a result of the retention ratio and capital adequacy ratio Granger cause the dividend payout ratio. Furthermore, the combined error of return on equity, capital adequacy ratio and dividend payout ratio Granger causes error of the retention ratio. The combined error of return on equity, retention ratio, and dividend payout ratio Granger causes error of the capital adequacy ratio and the combined error from return on equity, retention ratio, and capital adequacy ratio Granger causes error that occurs from the dividend payout ratio at 5 percent level of significance. In other words, knowing the combined error from the retention ratio and dividend payout ratio, the level of error from the capital adequacy ratio can be determined. The combined error from return on equity, retention ratio and capital adequacy ratio also determines the level of error from the dividend payout ratio. Similarly, following the study conducted by Dhamala et al. (2008) where it was established that causality can be tested using Fstatistics and probability values under the null hypothesis of no causality, the Pairwise causality test was used to test the causality between dividend policies and the ROE of SSA commercial banks as shown in Table 9 below, since none of the policies granger cause ROE in the estimate of the block exogeneity Wald test.
Table 9. Pairwise Granger Causality Test
Null Hypothesis: 
Obs 
FStatistic 
Probability 
Decision on Null Hypothesis 
RERA does not Granger Cause ROE 
2401 
0.23145 
0.79340 
Do not Reject 
ROE does not Granger Cause RERA 

0.96503 
0.38112 
Do not Reject 
CAR does not Granger Cause ROE 
2402 
0.84799 
0.42840 
Do not Reject 
ROE does not Granger Cause CAR 

3.22989 
0.03973** 
Reject 
DPOR does not Granger Cause ROE 
2363 
0.96910 
0.37958 
Do not Reject 
ROE does not Granger Cause DPOR 

3.60122 
0.02744** 
Reject 
CAR does not Granger Cause RERA 
2487 
1.18586 
0.30566 
Do not Reject 
RERA does not Granger Cause CAR 

4.36451 
0.01282*** 
Reject 
DPOR does not Granger Cause RERA 
2403 
8.44241 
0.00022*** 
Reject 
RERA does not Granger Cause DPOR 

20.3805 
1.7E09*** 
Reject 
DPOR does not Granger Cause CAR 
2404 
1.18027 
0.30737 
Do not Reject 
CAR does not Granger Cause DPOR 

2.57339 
0.04349** 
Reject 
Source: Author’s estimation, 2018
Note that *** and ** represents rejection of Ho at 1% and 5% respectively
The result of the Granger causality among the financial variables under consideration is presented in Table 9 to show the direction of causal relations between each pair of the financial variables such as ROE RERA, CAR and DPOR. The result shows that, there was a unidirectional causality between ROE and CAR; ROE and DPOR; RERA and CAR; and capital adequacy and DPOR and bidirectional causality between DPOR and RERA. This is evident from the estimated probability of Fstatistic values given as 0.039, 0.027, 0.013, 0.000, 0.002 and 0.004 ˂ 0.05.
Thus, i) ROE Granger causes capital adequacy and this implies that banks’ returns determine their ability to be adequately capitalized.
ii) ROE Granger causes DPOR, implying that the more banks earn from equity, the more they implement payout policy as against retention policy.
iii) RERA Granger causes capital adequacy and this implies that the more banks adopt dividend reinvestment plans, the more they generate earnings to increase their assets and solidify their capital base. This will not only serve as a cushion in times of shocks but promote their future growth.
5. Conclusions and Recommendations
Having established that RERA Granger causes bank performance using ROA and further conducting a robustness check using ROE as the measure of financial performance, our findings revealed that ROE Granger causes DPOR (unidirectional causality between ROE and DPOR) while neither DPOR nor RERA Granger cause ROE. This implies that when banks generate income from total shareholders’ equity, they will stick to payout policy even though this policy does not enhance their performance (ROE) in SSA. In conclusion, paying out does not create value because the unidirectional causality was from banks’ ROE to DPOR. A winlose game will result if banks continue to payout, as is the case with SSA banks. An optimal dividend policy that promotes the firm’s future growth must cater for future financing and increased assets. This finding is logical as what is generated should normally determine what will be paid out. However, given that SSA regional economic growth depends solely on the financial system and that the banking sector is at the forefront of the financial landscape, banks should adopt policy that will enhance growth.
6. References
Abdella, A.B. & Manual, V. (2016). A Study on the Determinants of Dividend Policies of Commercial Banks in Saudi Arabia. Imperial Journal of Interdisciplinary Research, 2(9), pp. 10641073.
Abiola, J.O. (2014). Measuring and Analyzing the Effects of Dividend Policy in Banking Profits and Growth. Journal of Policy and Development Studies, 9(1), pp. 167178.
Agyei, S.K. & MarfoYiadom, E. (2011). Dividend policy and bank performance in Ghana. International Journal of Economics and Finance, 3(4), pp. 202218.
Ajanthan, A. (2013). The relationship between dividend payout and firm profitability: A study of listed hotels and restaurant companies in Sri Lanka. International Journal of Scientific and Research Publications, 3(6), pp. 16.
Akande, J.O. & Kwenda, F. (2017). Competitive Condition of SubSaharan Africa Commercial Banks. Studia Universitatis BabesBolyai Oeconomica, 62(2), pp. 5576.
Akinlo, A.E. & Egbetunde, T. (2010). Financial development and economic growth: The experience of 10 subSaharan African countries revisited. The Review of Finance and Banking, 2(1), pp. 017028.
Al‐Twaijry, A.A. & Powers, M.R. (2007). Dividend policy and payout ratio: evidence from the Kuala Lumpur stock exchange. The Journal of Risk Finance, 8(4), pp. 349363. doi:10.1108/15265940710777306.
Allen, F.; Carletti, E.; Cull, R.; Qian, J.Q.; Senbet, L. & Valenzuela, P. (2014). The African financial development and financial inclusion gaps. Journal of African economies, 23(5), pp. 614642.
Allen, F.; Otchere, I. & Senbet, L.W. (2011). African financial systems: A review. Review of Development Finance, 1(2), pp. 79113.
Amidu, M. (2007). How does dividend policy affect performance of the firm on Ghana Stock Exchange? Investment Management & Financial Innovations, 4(2), pp. 103112.
Andriof, J.; Waddock, S.; Husted, B. & Rahman, S.S. (2017). Unfolding stakeholder thinking: theory, responsibility and engagement: Routledge.
Ashraf, B.N. & Zheng, C. (2015). Shareholder protection, creditor rights and bank dividend policies. China Finance Review International, 5(2), pp. 161186.
Beck, T. & Cull, R. (2013). Banking in Africa. World Bank Policy Research Working Paper No. 6684, pp. 144.
Beck, T.; DemirgüçKunt, A. & Maksimovic, V. (2004). Bank competition and access to finance: International evidence. Journal of Money, Credit and Banking, pp. 627648.
Brighi, P. & Venturelli, V. (2013). CEFIN Working Papers, No. 43, pp. 133.
Brighi, P. & Venturelli, V. (2014). How do income diversification, firm size and capital ratio affect performance? Evidence for bank holding companies. Applied Financial Economics, 24(21), pp. 13751392. doi:10.1080/09603107.2014.925064.
Bulan, L.T. & Subramanian, N. (2009). The firm life cycle theory of dividends. Dividends and dividend policy, pp. 201213.
Bulan, L.T.; Subramanian, N. & Tanlu, L. (2007). When Are Dividend Omissions Good News? pp. 139. doi: http://dx.doi.org/10.2139/ssrn.997427.
Byrne, J. & O’Connor, T. (2017). Creditor rights, culture and dividend payout policy. Journal of multinational financial management, 39, pp. 6077.
DeAngelo, H.; DeAngelo, L. & Stulz, R.M. (2006). Dividend policy and the earned/contributed capital mix: a test of the lifecycle theory. Journal of financial Economics, 81(2), pp. 227254.
Dhamala, M.; Rangarajan, G. & Ding, M. (2008). Estimating Granger causality from Fourier and wavelet transforms of time series data. Physical review letters, 100(1), 018701018730. doi: https://doi.org/10.1103/PhysRevLett.100.018701.
Díez Esteban, J.M. & López de Foronda Pérez, Ó. (2001). Dividend policy of European banks.
Ehikioya, B.I. (2015). An Empirical Investigation of the Impact of Dividend Policy on the Performance of Firms in Developing Economies: Evidence from Listed Firms in Nigeria. International Journal of Finance and Accounting, 4(5), pp. 245252.
Enoch, M.C.; Mathieu, M.P.H.; Mecagni, M.M. & Kriljenko, M.J.I.C. (2015). PanAfrican Banks: Opportunities and Challenges for CrossBorder Oversight: International Monetary Fund.
Fisher, A.M. (1993). Integer Cantor sets and an ordertwo ergodic theorem. Ergodic Theory and Dynamical Systems, 13(01), pp. 4564.
Flamini, V.; Schumacher, M.L. & McDonald, M.C.A. (2009). The determinants of commercial bank profitability in SubSaharan Africa: International Monetary Fund.
Francis, M.E. (2013). Determinants of commercial bank profitability in SubSaharan Africa. International Journal of Economics and Finance, 5(9), pp. 134147. doi:http://dx.doi.org/10.5539/ijef.v5n9p134.
Granger, C.W. (1969). Investigating causal relations by econometric models and crossspectral methods. Econometrica: Journal of the Econometric Society, pp. 424438.
Gujarati, D.N. & Porter, D.C. (2003). Basic Econometrics. 4th. New York: McGrawHill.
Gul, E. & Ekinci, A. (2006). The causal relationship between the real exchange rate and export and import in Turkey: 1990–2006. Journal of Social Sciences, 16, pp. 165183.
Hamid, R.; Yaqub, M.S. & Awan, M.M.H. (2016). Antecedents of Dividend Policy: Empirical Evidence from Banking Sector of Pakistan. European Online Journal of Natural and Social Sciences, 5(2), pp. 263278.
Houston, J.F.; Lin, C.; Lin, P. & Ma, Y. (2010). Creditor rights, information sharing, and bank risk taking. Journal of financial Economics, 96(3), pp. 485512.
Hsiao, C. (2014). Analysis of panel data: Cambridge university press.
Hyndman, R.J. & Athanasopoulos, G. (2014). Forecasting: principles and practice: OTexts.
Kablan, S. (2010). Banking efficiency and financial development in SubSaharan Africa: International Monetary Fund.
Ling, F.S.; Abdull Mutalip, M.; Shahrin, A.R. & Othman, M.S. (2007). Dividend policy: evidence from public listed companies in Malaysia.
Mecagni, M.; Marchettini, D. & Maino, R. (2015). Banking in subSaharan Africa: Key features and challenges. Recent Trends in Banking in subSaharan Africa, 9.
Mehran, H.; Morrison, A.D. & Shapiro, J.D. (2011). Corporate governance and banks: What have we learned from the financial crisis? FRB of New York Staff ReportNo. 502, 1(1), pp. 144. doi:https://dx.doi.org/10.2139/ssrn.1880009.
Mlachila, M.; Dykes, D.; Zajc, S.; Aithnard, P.H.; Beck, T.; Ncube, M. & Nelvin, O. (2013). Banking in subSaharan Africa: Challenges and opportunities. Retrieved from Mlachila, M.; Park, S. G. & Yabara, M. (2013a). Banking in SubSaharan Africa.
Mlachila, M.; Park, S.G. & Yabara, M. (2013b). Banking in subSaharan Africa: the macroeconomic context. Banking in subSaharan Africa, 1119. doi:https://www.econstor.eu/bitstream/10419/88938/1/776005626.pdf#page=12.
Moyo, J.; Nandwa, B.; Council, D.E.; Oduor, J. & Simpasa, A. (2014). Financial sector reforms, competition and banking system stability in SubSaharan Africa. New perspectives.
Mueller, D.C. (1972). A life cycle theory of the firm. The Journal of Industrial Economics, pp. 199219.
Nnadi, M.; Wogboroma, N. & Kabel, B. (2013). Determinants of dividend policy: Evidence from listed firms in the African stock exchanges. Panoeconomicus, 60(6), pp. 725741.
Obembe, O.; Imafidon, J. & Adegboye, A. (2014). Product Market Competition and Dividend Payouts of Listed NonFinancial Firms in Nigeria. International Journal of Economics and Finance, 6(11), pp. 117130.
Odunga, R.; Nyangweso, P.; Carter, D. & Mwarumba, M. (2013). Credit Risk, Capital Adequacy and Operating Efficiency Of Commercial Banks in Kenya. International Journal of Business and Management Invention, 2(9), pp. 612.
Onanjiri, R. & Korankye, T. (2014). Dividend Payout and Performance of Quoted Manufacturing Firms in Ghana. Research Journal of Finance and Accounting, 5(15), pp. 442454.
Patra, T.; Poshakwale, S. & OwYong, K. (2012). Determinants of corporate dividend policy in Greece. Applied Financial Economics, 22(13), pp. 10791087.
Priya, K. & Nimalathasan, B. (2013). Dividend Policy Ratios and Firm Performance: a case study of Selected Hotels & Restaurants in Sri Lanka. Global Journal of Commerce and Management Perspectives, 2(6), pp. 1622.
Shao, L.; Kwok, C.C. & Guedhami, O. (2013). DIVIDEND POLICY: BALANCING SHAREHOLDERS'AND CREDITORS'INTERESTS. Journal of financial Research, 36(1), pp. 4366.
Thanatawee, Y. (2011). Lifecycle theory and free cash flow hypothesis: Evidence from dividend policy in Thailand.
Uwuigbe, U.; Jafaru, J. & Ajayi, A. (2012). Dividend policy and firm performance: A study of listed firms in Nigeria. Accounting and Management Information Systems, 11(3), pp. 442455.
Waseem, A.; Saleh, A.; Shukairi, M. & Mahmood, N. (2011). The effect of dividend policy stability on the performance of banking sector listed on Amman Stock Exchange. International Journal of Humanities and Social Science, 1, pp. 201205.
Zakaria, Z.; Muhammad, J. & Zulkifli, A.H. (2012). The impact of dividend policy on the share price volatility: Malaysian construction and material companies. International Journal of Economics and Management Sciences, 2(5), pp. 18.
Zhou, P. & Ruland, W. (2006). Dividend payout and future earnings growth. Financial Analysts Journal, 62(3), pp. 5869.
1 Department of Management Accounting, Faculty of Accounting and Informatics, Durban University of Technology, South Africa, Address: Ritson Campus, 7 Ritson Road, Durban, South Africa, Corresponding author: OdunayoO@dut.ac.za.
AUDŒ, Vol. 14, no. 3, pp. 254277
Refbacks
 There are currently no refbacks.
This work is licensed under a Creative Commons Attribution 4.0 International License.