ABSTRACT

This study seeks to explore the link between Corporate Social Responsibility (CSR) disclosure and Financial Performance (FP) in the banking industry.  To test our predictions, we use a sample from the Mozambican banking sector over the period from 2009 to 2017. Applying Return On Equity and Return On Asset as a measure of FP and, content analysis to assess the degree of CSR engagement. Our results indicate a significant and positive association between CSR disclosure and FP, implying that the Mozambican banking sector is socially responsible and, that its involvement in CSR disclosure enhances its FP. Secondly, we find a bidirectional association between FP and CSR. Furthermore, we find that all CSR disclosure dimension has an impact on the FP.

KEYWORDS: Corporate Social Responsibility, Financial Performance, Banking Sector

INTRODUCTION

To date, companies should not only maximize their shareholders’ return but also their activities should benefit the whole society. Companies must be managed and conducted in a transparent manner, with great respect to the environment, community and employees.

Waddock and Graves (1997) argue that a company should not only provide CSR activity but should provide and report it because it will enhance its reputation among its key stakeholders and consequently achieve the best financial performance. CSR disclosure is the main characteristic of CSR firms and is a signal of the company`s value by investors. Following the arguments of Siueia and Wang (2017) the banking sector is vital for a country’s economy and the whole society. Consequently, researchers are growing number and number of empirical research in CSR activity and disclosure in the banking industry (Bae, Kang and Wang, 2011; Goss and Roberts, 2011; Wu and Shen, 2013; Esteban-Sanchez et al., 2017; Platonova et al., 2016).

This study seeks to explore the link between Corporate Social Responsibility (CSR) disclosure and Financial Performance (FP) in the Mozambican Banking industry over the period from 2009 to 2017. Thus, this study tries to answer these two questions: “is there a significant and positive relationship between CSR and FP in the Mozambican Banking industry? And whether the implementation of CSR disclosure behavior conduct a high degree of financial performance in the Mozambican Banking sector and vice-versa?” To answer the questions, we following the accounting literature. In the literature, we can find the different type of theory that examines the relationship between CSR and FP (Stakeholder theory, legitimacy theory, and agency theory). To solve the research questions, we applied the stakeholder theory which, in our viewpoint, better explain this association.

We test our hypotheses using a dataset extracted from the audited annual statement of Mozambican banks available online at the Bank of Mozambique (Central bank of Mozambique), Mozambican association of banks and individual banks’ website. Following earlier studies in this field Esteban-Sanchez et al. (2017); Platonova et al. (2016); Flammer (2015); Simpson and Kohers (2002); Waddock and Graves (1997), we used the appropriate measure of FP based on profitability, such as ROE (return on equity) and ROA (return on asset). In addition, we applied the content analysis to assess the CSR disclosure index extracted from bank annual reports and extracted from the individual bank website (Dyduch and Krasodomska, 2017; Platonova et al., 2016; Branco and Rodrigues, 2006). We control for a possible causal relationship between CSR disclosure and FP by using panel data regressions fixed effects model and, we performed two-stage least square (2SLS) analysis, in order to solve the endogeneity problem.

The main findings of our study are as follows. Firstly, we found a significant and positive association between FP and CSR implying that the Mozambican banking sector is socially responsible and, that a bank`s engagement to CSR disclosure enhances better FP. Secondly, we found a bi-directional association between FP and CSR, suggesting that CSR disclosure behavior accrue higher profitability to the banks. Therefore, this study highlights that customer & products, local community, and human resource policies can drive bank managers to generate better financial performance, however environmental protection initiative have an insignificant impact on FP.

This paper will be focused on Mozambican banks due to the following reasons: Mozambique is a representative member of the Southern African Development Community (SADC) and, during the years 2015-2017, Mozambique underwent a hidden public debt crisis, making this country an exclusive one among SADC and, particularly an unfavorable country to invest (Siueia and Wang, 2017), provides a compelling scenario for examining the link between CSRD and FP, for countries where such a study still needs to be done, particularly in the African scenario. On the other hand, foreign banks operating in Mozambique also operate SADC and it is crucial that banks adopt a strategic approach to CSR disclosure in order to remain competitive in financial business. According to Porta and Kramer (2006), the location of the company is one of the condition to understand the company CSR policy. Therefore, Esteban-Sanchez et al. (2017) argue that these previous findings cannot be generalized to different contexts, and CSR activities vary across counties.

Our study attempts to fill the existing gap in the accounting literature, contributing to the knowledge of how banks strive to achieve CSR disclosure behavior and improve better FP in underdeveloped countries, particularly in Mozambique (an example of sub-Saharan African countries). To the best of our acknowledge, empirical studies that examined the impact of CSR disclosure on financial performance in sub-Saharan African countries, including Mozambique, are almost non-existent and, this is the first empirical evidence that attempts to examine the unexplored bidirectional link between CSR disclosure and FP in the banking sector. We further continue to clarify the link between CSR disclosure and FP which could shed new light on the CSR`s behavior in the Mozambican banking sector. Therefore, the evidence from the study can help policy-makers, regulators, and investors to understand banks’ business practices in Mozambique.

LITERATURE AND HYPOTHESES

Following previous research such as Esteban-Sanchez et al. (2017); Platonova et al. (2016); Flammer (2015); Simpson and Kohers (2002); Waddock and Graves (1997), CSR has been defined in several ways and over time, its contents are also evolving. However, we use the widely acceptable definitions of Jain et al. (2015). For Jain et al. (2015), CSR can be defined as “the company’s commitment to contributing to sustainability initiatives, such as developing good relationships with employees, the local community and society at large to improve quality of life, which is good for business and good for sustainable development”. Given the impending challenges of the banking sector, banks must support the activities of corporations and organizations that are on a positive plane to ensure that society develops and meet their expectations, for this they need to actively engage in CSR activity and disclosure (Esteban-Sanchez et al., 2017).

To develop our hypotheses to examine the bidirectional link between CSR disclosure and FP, we rely on prior studies, as shown in figure 1.

[Figure 1]

CSR disclosure is the main characteristic of CSR firms and is a signal of company`s quality by investors, Cox and Wicks (2011) argue that CSR disclosure is an effective way for them to obtain a report about CSR behavior. Dhaliwal et al. (2014) believe that the disclosure of CSR is an essential complement to financial reporting and could influence investor behavior. Along the same line, Cheng et al. (2013) and Cormier et al. (2011) argues that investors have difficulty accessing a reliable financial report, the disclosure of CSR is seen as an important tool for management’s credibility, honesty, and reliability. According to previous studies (Dyduch and Krasodomska, 2017; Platonova et al., 2016; Branco and Rodrigues, 2006), CSR behavior is subjective and Content analysis is a credible proxy for the measurement of socially responsible performance. On the other hand, several theories explain the positive link between CSR-FP, however, three of them seem to prevail: agency theory, legitimacy theory, and Stakeholder theory (Dyduch and Krasodomska, 2017; Manrique and Martí-Ballester, 2017; Esteban-Sanchez et al., 2017; Platonova et al., 2016). In our analysis, we consider stakeholder theory to solve the research questions, in our viewpoint, this theory better explains positively the link between these two variables. Stakeholder theory suggested the idea that the companies that adopt socially responsible behavior are building their reputation in the society, attracting more investors to invest in the company, consequently reducing the operational costs, generating their competitive advantages and increasing their profitability. (Surroca et al., 2010; Soana, 2011; Goss and Roberts, 2011; Wu and Shen, 2013; Platonova et al., 2016; Manrique and Martí-Ballester, 2017; Dyduch and Krasodomska, 2017).

While the CSR disclosure differs from company to company, there is no doubt that this is characterized by voluntary disclosure, generate transparency and, is a great tool for business and CSR disclosure firm are more transparent relative to non-CSR disclosure firm. Following Sholtens (2009) particularly in the banking industry, society is demanding that managers increase their commitment to socially responsible action, being more transparent and become more involved in CSR disclosure, as these actions increase the bank`s reputation and profitability. Further, the highest level of the company`s CSR disclosure index, the greater the level of FP.

The study of Javed et al. (2016) conclude that there are no clear conclusions about the link between CSR-FP; however, the majority reported a positive, other negative or, inexistent or, mixed correlation. Nevertheless, to the best of our knowledge, there are few prior researches concerning the link in the banking sector. Empirically, the research of Goss and Roberts (2011); Wu and Shen (2013); Platonova et al. (2016); Esteban-Sanchez et al. (2017), concluded that CSR is positively associated with profitability and, banks are involved in CSR to improve their financial performance. Although these studies have been found for western countries, both developed and developing countries, it may reflect the same in underdeveloped countries like Mozambique. However, the bidirectional CSR-FP association has not been studied in the banking sector. Considering the stakeholder theory suggests that all business should adopt socially responsible behavior, as they generate high profitability, we could predict higher levels CSR disclosure index to be positively linked to greater levels of FP and vice versa. In general, we predict bank managers to have the incentive to engage in CSR disclosure to archive better financial performance and the FP will influence CSR disclosure. Based on the theoretical and empirical justifications above, our hypotheses are presented as:

H₁: Ceteris paribus, a higher degree of CSR disclosure influence greater banks’ profitability (Model 1).

H₂: Ceteris paribus, better profitable banks are those that adopt greater CSR disclosure, (Model 2).

RESEARCH DESIGN

Data and sample

This study uses bank dataset extracted from the annual reports of Mozambican banks available online at the Bank of Mozambique (Central Bank of Mozambique-BM), Mozambican association of banks and individual banks’ website.

In Mozambique by the end of 2017, the banking sector comprised only 19 banks, with the majority being foreign banks (BM). Though, all banks in Mozambique observe the same banking regulations and are under the supervision of the bank of Mozambique. However, they have different CSR policies. With this, perhaps the CSR strategies carried out among the banks may also be different. We excluded in our sample the Central Bank of Mozambique because of their different characteristic (currency-issuing bank) compared to other banks. Considering the small size of banks in Mozambique, and not to lose further observations, we did not exclude foreign banks. Since it was difficult to access databases of the banks, we restricted our sample to 18 banks, with at least eight years of dataset available, spanning 2009 to 2017. Finally, in order to minimize the effect of outliers in the regressions, the upper and lower extreme values in each variable were disqualified (1% standard deviations).

METHODOLOGY

Measurement of CSR

As noted previously, several studies used different methods to measure CSR behavior (one-dimensional indicators, ethical rating, reputational measures, surveys, content analysis). In line with previous studies, we applied content analysis approach to measure the level of CSR disclosure activity. (Branco and Rodrigues, 2006; Dyduch and Krasodomska, 2017; Platonova et al., 2016). Content analysis is a research technique used to process, examine, analyze, interpret, and classify different types of content data and permit to segregate CSR practices into different categories. Thus, we carefully collected, processed and analyzed the total number of statements provided to describe each category of social information for each bank. More specifically the data were extracted from the annual reports of banks and their levels of corporate social disclosure, were obtained by summing all the four key stakeholder components (costumer & products information, Human resources information, local community involvement information, and environmental initiative information). More specifically, each of these categories was divided into sub-categories of twenty-two CSR activities as observed in (Table 1). We supplement these data searching the banks’ websites to identify banks which have issued CSR in their reports. Banks were considered as disclosing banks if there was CSR report on their website and non-disclosing banks where otherwise (Dyduch and Krasodomska, 2017; Platonova et al., 2016). CSR reports through their website as a stand-alone document for the June of 2016 to October of 2016 and from July of 2017 to December 2017. If the disclosure item seems to be many times or in two or more report it was only treated once. Finally, we measured the CSR disclosure (CSR_Din) as a sum of CSR disclosure score from the annual report and the bank`s website report. The CSR disclosure index (CSR_Dindex) was released nine consecutive years and the results obtained from bank disclosed information in its annual reports and on its websites. In accordance with the research of Platonova et al. (2016), we estimate the CSR disclosure index using the following equation:

CSR_Dindex=∑i=1nKijtn

(1)

Where CSR_Dindex_jt indicates the CSR disclosure index for dimension j and period t; K_ijt is variable K (ranging from 1 to n) for dimension j and period t; n is the number of statements. It should be noted that the adoption of this model arises because in Mozambique there are no rating organizations – trusted third-party institutions with the resources to collect the necessary information for quantitative CSR disclosure.

Measurement of FP

Earlier studies used different methods to measure financial performance (return on equity, return on assets, the growth rate of sales, return on sales, others studies used market measures such as market to book ratio or stock returns and,  net profit margin. (Esteban-Sanchez et al., 2017; Platonova et al., 2016; Flammer, 2015; Simpson and Kohers, 2002; Waddock and Graves, 1997). In line with Flammer (2015), Platonova et al. (2016) and Esteban-Sanchez et al. (2017), we applied ROA and ROE as the measure of FP.

Baseline Model

To test the hypotheses first, the equations of the models were estimated through panel data regressions applying a fixed effects model. Further, for robustness check, we performed two-stage least square (2SLS) analysis, in order to check endogeneity problem, following prior study. Specifically, we formulate our baseline models as follows:

We propose equation (2) to test the effect of CSR on FP H₁. If H₁ hold true we expect the coefficient of CSR_Dindex_j,t (β₁) to be positive and higher in magnitude and significance.

FP=β0 + β1CSRDindex+ ∑k=1Kλi*Zi,t +δt+ θi + εi,t

(2)

Where all variables are defined in Table 2. FP is a measure of Financial Performance based on ROE and ROA; CSR_Dindex comes from Eq. (1); Z_i,t representing a series of the control variable, namely bank size (LNSize), the capital ratio (Cap­_R), Loan ratio (Loan_R) and, Debt ratio (Debit_ R). Finally, β₀ indicates the intercept; β₁ and λ_i designate the regression coefficients; δ_t donates time-fixed effects; θ_i denote unobservable bank-specific effects as bank corporate culture. And ε_i,t denoted the error term.

Furthermore, to examine whether the high FP have an impact on their socially responsible disclosure in line with the possible bidirectional link between these two variables, that was found in non-financial sector studies (Rodriguez-Fernandez, 2016), we propose the Eq. (3), model 2, to test our H₂. If H₂ hold true we expect the coefficient of FP (β₁) to be positive and higher in magnitude and significance.

CSR_Dindex=β0 + β1FP+ ∑k=1Kλi*Ni,t +δt+θi + εi,t

(3)

Where all variables already are defined.

Variable Explanation

Dependent Variable

To measure the CSR disclosure we used content analysis following prior researches (Dyduch and Krasodomska, 2017; Platonova et al., 2016), and to measure bank FP we use widely profitability measures such as ROE and ROA. Following prior research (Esteban-Sanchez et al., 2017; Platonova et al., 2016; Flammer, 2015; Simpson and Kohers, 2002), we exclude the measure of market-return in this study because we believe that some of the banks have not been listed on stock exchanges. CSR_Dindex comes from Eq. (1) is used as dependent in concordance with the possible bidirectional relationship between CSR and FP (Rodriguez-Fernandez, 2016).

Independent Variable

FP and CSR_Dindex were previously explained and used as independent variable in concordance with the possible bidirectional relationship between CSR and FP (Rodriguez-Fernandez, 2016).

Control Variable

Reference to prior studies we added several control variables that could affect CSR disclosure and financial performance. First, we controlled the Firm size (Size) measured by the natural logarithm of total assets. Size has widely been recognized as a proxy for better profitability. The larger bank size allows it to attract cheaper capital, thus, the bank will have more recourse to invest in CSR activities. The larger bank makes more contributions to the whole society comparatively to the small banks (Wu and Shen, 2013). We expect that the relationship between firm size and FP will be positive. Second, in the same line, we control another proxy of the profitability ratio, the capital ratio measured as equity capital divided by total assets. It’s an indicator of bank invisible risk of default that indicates the bank’s ability to grow under the present capital structure. Patonova et al. (2016) argued that the highest capitalized banks tend to engage aggressively in CSR behavior. The expectation is a positive and significant association between CSR and FP. Third, we included Loan ratio calculated as total loans divided by total assets to control the earning power of the bank (Simpson and Kohers 2002). This variable is used as an indicator of bank credit risk and also represent one of the main sources of banks’ revenue. In agreement with Siueia & Wang (2017), bank managers can increase provision when they presumed that loan will decrease and affect negatively the profit, thus Loan ratio is expected to have a positive relationship with FP and consequently with CSR. Debt reduces the profitability at the same time, debt is also a frequently tested factor to analyze the influential elements of CSR disclosure policies (Barnett and Salomon, 2012). It’s expected that the coefficient on the debt ratio variable will be negative through the FP and CSR.

STATISTICS AND CORRELATION

Table 3 summarizes the descriptive statistics of the variables of interest for the object of the research. In terms of the variable CSR_Dindex that represent the CSR disclosure index, we can observe the minimum and maximum values ranging between 3% and 64% with a mean value of 0.4398 and a standard deviation of 0.1998. This implies, that more than 43% of banks have reported voluntary CSR in their annual report or their bank website during the scope of our study, these results also, suggest less engagement in CSR disclosure consequently the banks did not perform their CSR disclosure. In terms of the performance variable, we can observe that the highest value for ROA is 14% and for ROE 17%. The respective standard deviation is 0.2016 and 0.1857, with the mean 2.84% and 3.07% respectively. However, we underscore the period under study and indicate that banks recorded losses according to the usual negative outcome of the ratio of FP. In terms of our control variables, we can observe that all were as predicted.

Also, a Pearson correlation matrix of the variables used in our study is presented in Table 3. We drew attention to the correlation between each FP variables and CSR disclosure index, and be observed that there was a positive and significant correlated at the conventional level. These results imply that the Mozambican banking industry is socially responsible and consequently supported our hypothesis H₁, confirming the prior researches in the field (Simpson and Kohers, 2002; Wu and Shen, 2013; Platonova et al., 2016). The results also displayed the highest VIF – Variance Inflation Factor value as 2.422 with a mean of 1.201 for the CSR_Dindex variable, suggesting that the multi-collinearity does not appear to be a problem.

[Table 3]

EMPIRICAL RESULTS

The results applying the regression in model 1 are presented in Table 4. This model investigates the effects of voluntary CSR disclosure on financial performance. The first column, display the estimations for equation (2) using ROA as a measure of FP, the last column displays estimations for equation (2) using ROE as a measure of FP. The equations of the models were estimated through panel data regressions applying a fixed effects model. We carried out the Hausman’s test (Hausman and Taylor, 1981) and the null hypothesis is rejected (p-value < 0.000; significant at 5%), suggesting that using the fixed effects model was the most accurate estimation.

In terms of heteroscedasticity, we used the modified Wald test to solve it, as can be seen in Table 4, the value of the modified Wald test is at the conventional level, we can validate the null hypothesis, inferring no threat of heteroscedasticity. We used the Wooldridge (A1) test to identify autocorrelation, as shown in Table 4, the autocorrelation does not appear to be a problem because our highest value of Wooldridge (A1) test was 2.241, so we reject the residual autocorrelation hypothesis. This result, also displayed the highest VIF – variance inflation factor value as 2.453 with a mean of 1.204 for a CSR_Dindex variable, suggesting that the multi-collinearity does not appear to be a problem.

The ROA model, Adj. R-squared equal 37%; indicating the power of our model. We can observe the greater magnitude and significance of the coefficient on CSR_Dind_j,t (β₁ = 0.0693 p-value < 0.05 or better, t-stat = 3.281), demonstrating that CSR disclosure behaviors have a positive influence on FP, suggesting that banks increase their FP level when CSR disclosure rises. In term of our control variables, it can be observed that all coefficients are statistically significant at the conventional level. Particularly, for size variable “LNsize” (λ₁ = 0.0513, P-value <0.05 and t-stat = 3.214), revealing that large banks are more engage in CSR disclosure supporting the augments of Wu and Shen (2013), Platonova et al. (2016). Second, we control the capital ratio (λ₁ = 0.0462, P-value <0.05 and t-stat = 3.001), suggesting high growth banks are more likely to engage in CSR disclosure in support of augments by Simpson and Kohers (2002), Patonova et al. (2016). The coefficient of loan ratio is λ₁ = 1.982, P-value <0.05 and t-stat = 3.223), revealing that our sample has less credit risk (Platonova et al., 2016; Simpson and Kohers 2002). On debt ratio the coefficient is negative and significant as expected, revealing that our model is consistent with earlier studies (Barnett and Salomon, 2012).

Likewise, concerning the ROE model (first robustness test). As presented in Table 4, we can observe that the results are very similar to the ROA model. Especially, the CSR disclosure indexes coefficient is still positive and significant at the conventional level (β₁ = 0.0499 and t-stat = 3.017) suggesting that CSR disclosure behaviors still have a positive impact on FP, implying that banks increase their FP level when CSR disclosure raises.

Regarding our four control variables, it was seen that all were unchanged and still statistically significant at the conventional level as predicted. Additionally, the Adj. R-squared is equal to 0.395 means that our model explains about 40% of the behavior of the independent variable. The results confirm the expectation based on the stakeholder theory and also support our hypothesis H₁, Suggesting that voluntary CSR disclosure is positively linked to the FP of banks in the Mozambican banking sector. Therefore, these findings can be interpreted as being a higher level of CSR disclosure, and better bank’s performance. Our findings are in line with Platonova et al. (2016), Wu and Shen (2013), Scholtens (2009), Simpson and Kohers (2002) studies and oppose the findings of Bae, Kang, and Wang (2011).

[Table 4]

Testing our second hypothesis, we examine model 2, which investigate the impact of FP on voluntary CSR disclosure index. The results are shown in Table 5. The first column, display the estimations for Eq. (3) using ROA as a measure of FP, the last               column displays estimations for Eq. (3) using ROE as a measure of FP. As can be seen, all the models are significant at least at the conventional level.

The equations of the models were estimated through panel data regressions applying a fixed effects model. We carried out the Hausman’s test (Hausman and Taylor, 1981) and the null hypothesis is rejected (p-value < 0.000; significant at 5%), suggesting that using the fixed effects model was the most accurate estimation. In terms of heteroscedasticity, we used the modified Wald test to solve it, as presented in Table 5, the value of the modified Wald test is at the conventional level, we can validate the null hypothesis, inferring no threat of heteroscedasticity. We used the Wooldridge (A1) test to identify autocorrelation, as observed in Table 5, the autocorrelation does not appear to be a problem because our highest value of Wooldridge (A1) test was 2.304, so we reject the residual autocorrelation hypothesis. This result, also displayed the highest VIF value as 2.292 with a mean of 1.217, suggesting that the multi-collinearity does not appear to be a problem. The highest Adj. R-squared equal 39%.

We highlighted the ROA variable, as shown is positive and statistically significant at the conventional level (β₁ = 0.0085 and t-stat = 3.116), revealing a positive effect on CSR disclosure. Regarding our four control variables, it can be observed that all are statistically significant at the conventional level as expected, suggesting that our model is consistent. According to the results, we can we can affirm our hypothesis H₂ because it is they in agreement with our prediction that there is a positive link between FP and CSR disclosure in the Mozambican banking sector. Therefore, we can state that the better level of FP, the greater the level of voluntary CSR disclosure. Similarly, the findings support the bidirectional relationship between CSR and FP and vice versa, according to the non-financial sector study of Rodriguez-Fernandez (2015).

Likewise, regarding the ROE model, as presented in Table 5, we can observe that all the results were unchanged and still support our hypothesis H₂. However, our models did not present multi-collinearity problems, autocorrelation problem even the problems of heteroscedasticity because we carried out the respective statistics test to solve it.

[Table 5]

ADDITIONAL ANALYSIS

Additionally, we carried out analyses to address possible alternative justifications for our results. We re-run our model applying each individual category of voluntary CSR disclosure CSR_Dis1 = customer and products, CSR_Dis2 = local community, CSR_Dis3 = environmental protection, CSR_Dis4 = human resource policies) as metrics of voluntary CSR disclosure score on FP measure (ROA and ROE). As presented in Table 6, Regarding the ROA model, all individual voluntary CSR disclosure is positively related to FP at the conventional level, the results indicate that these categories of voluntary CSR disclosure does seem to have positive impact on FP, suggesting that the banks use these elements of social responsibility behavior as tools to increase FP, except for environmental protection initiative category (CSR_Dis3) that the coefficient is statistically insignificant at conventional level (p-value > 10%). This means that the banks disclosure less in the environmental report category comparatively to another category of CSR disclosure present in our research. The insignificant impact of the environmental initiative ((λ₁ = 0.001, P-value > 0.01) on ROA, is maybe because banks do not produce direct goods and are not considered particularly polluting the environment, so they neglect the environment initiative and therefore do not disclose information concerning this category of voluntary CSR disclosure. (Scholtens and Dam, 2007 in Esteban-Sanchez et al, 2017). On the other hand, this finding can be explained by the fact that in non-developed countries, bank managers are not interested in pursuing environmental issues as opposed to developed countries where manager focus their CSR efforts on environmental issues and issues that contribute to the development of the community.

Regarding our control variables, it was seen that all were unchanged and still statistically significant at the conventional level as expected. Additionally, the highest Adj. R-squared was 32%. Also, we carried out the Hausman`s test for our model and the null hypothesis is rejected (p-value < 0.000; significant at 5%), further, we checked the multi-collinearity carried out by the VIF and the highest value was 2.399 implying that for our models the multi-collinearity does not appear to be a problem. Therefore, we carried out the Wooldridge (A1) test to check the autocorrelation problems. Similarly, the result reported in table 4, our highest value of Wooldridge (A1) test was 2.402, so we reject the residual autocorrelation hypothesis.  We used the modified Wald test to identify heteroscedasticity, the result presented in Table 5, validated the null hypothesis, inferring no threat of heteroscedasticity.

Regarding the ROE model, the findings were consistent with those presented in the ROA model, however, our models did not present nether multi-collinearity and autocorrelation problems, nor even the problems of heteroscedasticity because we carried out the respective statistics test to solve it, as shown in Table 6.

[Table 6]

INSTRUMENTAL VARIABLE (2SLS) ANALYSIS

Followingprevious studies, to avoid endogeneity problems we carried our sensitive test to validate the interpretation of the results. We use positive CSR disclosure score (CSR_DPos) as an alternative measure of voluntary CSR disclosure index. As shown in Table 7, the instrumental variable techniques using both model ROA and ROE. The coefficient of CSR_DPos is still positive and significant at the conventional level for both models (λ₁ = 1.066, P < 5%) similar to those obtained in Table 5. Further, we applied the Durbin-Wu-Hausman test to check endogeneity, as presented in Table 7 (λ= 58.0121), the result implies that endogeneity cannot account for the positive and significant relationship between CSR disclosure and financial performance. Similarly, we carried out the F-test across for the two models were 33.06 and are above the conventional level (minimum of 10) suggesting that our instrument is robust. Hence, we rejected the null hypothesis (p < 0.000).

[Table 7]

CONCLUSION

An empirical study was undertaken to examine the relationship between CSR and FP, using the Mozambican banking industry as a sample covering the period from 2009 to 2017. We found that there is a positive and statistically significant effect of voluntary CSR disclosure on financial performance, also we found a bidirectional relationship between CSR disclosure and FP, suggesting that and CSR disclosure behavior leads to increased profitability of banks and the higher profitability is driven by CSR disclosure, thus, create a positive and significant relations in both directions. These findings support our hypotheses. It is evident that banks obtain economic benefits from their voluntary CSR disclosure, thus bank managers should engage more on voluntary CSR disclosure.

Nevertheless, we believe that our study is a novelty, the first studying the bidirectional relationship between CSR disclosure and FP in the banking sector because in our literature review there was scanty evidence. Second, in the Mozambican banking industry this research is a novelty, because of our literature review there was scanty evidence.

Future research could consider the application of a different measure of CSR disclosure and expand the period of analysis. We believe that this research will contribute to filling the existing gap in the literature. Also, highlighting the importance of the manager engaged in CSR disclosure to achieve a better FP and confirm the stakeholder theory in the banking sector, particularly in the Mozambican banking sector.