Theresults applying the regression in model 1 are presented in Table 3.
This modelinvestigates the effects of voluntary CSR activities on financial performance.The first 3 columns, display the estimations for equation (2) using ROA as ameasure of FP, following the last 3 columns displays estimations for equation(2) using ROE as a measure of FP. Our model overall is statically significant(P-value <0.01) and the model for ROA shows Adj. R-squared equal42.2%; and for ROE the Adj. R-squared is 20.
1% respectively, indicatingthe strength power explanatory of our model. This result, also displayed thehighest VIF – variance inflation factor value as 2.481 with a mean of 1.211 fora CSR_Dindex variable, suggesting thatthe multi-collinearity does not appear to be a problem. On the other hand, alsothe autocorrelation does not appear to be a problem because our highest valueof Durbin-Watson test was 2.
102, so we reject the residual autocorrelationhypothesis. Also, we included the AR (1) term. To solve the endogeneity problemwe carried out by the Hausman’s test (Hausman and Taylor, 1981) for all models. Particularly, for model 1 the first 3 columns inTable 3, it can be observed that the coefficient ?1 on CSR_Dindexvariable is significantly positive (?1 = 0.0581, P-value <0.01and t-stat = 1.240), for the ROE measure of FP it can be observed that similarthe coefficient ?1 on CSR_Dindex variable is significantly positive(?1 = 0.0049, P-value <0.
01and t-stat = 2.398, respectively). Regarding our four control variables, it can beobserved that all are statistically significant (at 0.
1, 0.05 and 0.01respectively) and positive as excepted, debt ratio variable is negativesuggesting that our models are consistent. Jointly the result implies thepresence of voluntary CSR activities in FP behavior. The findings confirm theexpectation based on the stakeholder’s theory, and led to support our hypothesis.H0 and, H1.
Suggesting that Mozambican banking industry are voluntary socially responsible and with this bank behaviorincreases their financial performance. Our findings are in line Wu and Shen (2013); Platonova et al. (2016) studiesand opposes the findings of Bae, Kang, and Wang (2011);Kamal (2013) studies.(Insert around here Table 3)Regarding theregression in model 2, that investigate the impact of each categories ofvoluntary CSR activities (CSR_D1 = customer and products, CSR_D2 = localcommunity, CSR_D3 = environmental protection, CSR_D4 = human resource policies)on the FP measure (ROA and ROE) we can observe in Table 4, that each individualvoluntary CSR activity are positively related with FP, suggesting that thebanks use this for elements of social responsibility behavior as a tools toincrease FP, except for environmental protection initiative category (CSR_D3)it means that the banks invest less in the environmental protection activitycomparatively to another category of CSRactivities present in our research. The insignificant impact of environmentalinitiative ((?1 = 0.001, P-value > 0.01) on FP, is maybe becausethe banks don’t produce direct goods, so they careless about the environmentand consecutively do not disclose information concerning this category ofvoluntary CSR activity. Or on the other hand, in agreement with the study ofKvasi?, Cerovi? and Drazenovi? (2016); this canbe explained by the fact that in non-developed countries, bank managers are notinterested in pursuing the environmental issues as opposed to developedcountries where managers focus their CSR efforts on environmental issues andissues that contribute to the development of the community.
Thesefindings, also suggest that the individual element of voluntary CSR activity donot affect the FP. Therefore, this result indicates no statisticallysignificant association between the financial performance and all theindividual categories of CSR behavior, implying no sufficient evidence tosupport our hypothesis H2. On the other hand, wecan draw attention to the highest significant, and positive coefficient oncustomer and products variable (?1 = 0.
6271, P-value < 0.05 andt-stat = 2.011) implying that the banks are disclosing firstly information relative to their costumer and products,ignoring other categories of voluntary CSR behavior. This means that bankmanagers mostly disclose information related to customer and product. Thishappens because the bank managers do not have enough knowledge about theadvantage to participle on all individual categories of voluntary CSR behavior.
On the other hand, maybe because Mozambique is an under-developed country andthe bank managers do not have an incentive to participate in full CSRactivities. Because, in concordance with Kvasi?, Cerovi? and Drazenovi? (2016); the levelof development of the country is fundamental to determine the CSR practices ofthe bank managers. On the other hand, the bank managers are mostly involved inCSR behavior via costumer & product category, maybe they need to increasethe fidelity of their customers, influence new customers and create newrelations with the customer to promoting new deposits, which will have a greateffect on the banks’ financial performance.
These results are similar to theinternational research of Kvasi?,Cerovi? and Drazenovi? (2016); Kiliç (2016) and opposes the findings of Botshabelo et al. (2017). This results supports our hypothesis H2a,implying that the Mozambican Banking industry primarily disclose informationregarding their costumer and products and services.
Generally,our model 3 is significant and explanatory as presented in (Table 4), for themodels of ROA (Adj. R-squared 53.4%), and ROE (Adj. R-squared28.6%). Also, we carried out the Hausman`s test (Hausman and Taylor, 1981) for all models to solvethe endogeneity problem, additional, we checked the multi-collinearity carriedout by the VIF – Variance Inflation Factor statistic and the highest value was2.231 implying that for our models the multi-collinearity does not appear to bea problem.
Therefore, we carried out the Durbin-Watson statistic to check theautocorrelation problems, similarly the result presented in table 3, ourhighest value of Durbin-Watson was 1.787, so we reject the residualautocorrelation hypothesis.(Insert
