Two contradictory hypotheses emerge when we consider the effect of CSP on FP. The ‘social impact hypothesis’ – which assumes that meeting the needs and expectations of various stakeholders increases firm value – and, secondly, the ‘trade-off hypothesis’, which assumes that CSP increases results in firms incurring unnecessary costs which so lead to a reduction in value (Donaldson & Preston, 1995). This puts socially proactive and responsible firms at a disadvantage compared with less socially active ones (Peng & Yang 2014; Preston & O’Bannon 1997). The CSP and FP nexus has extensively garnered prior research, with some widely cited groundbreaking studies (e.g. McGuire et al. 1988; Orlitzky et al. 2003). Nevertheless, to date the available empirical findings remain largely mixed and inconsistent (McGuire et al. 1988). Advocates of a positive CSP–FP nexus hold a view that not only may voluntary CSP be considered a business objective in itself, but it also serves as a means to an end with regard to FP. On the other hand, those who found a negative relationship suggest that firms’ investment in voluntary CSP diverts funds that could have been channelled into other productive investments (Peng & Yang 2014; Preston & O’Bannon 1997). As argued in extant literature, the mixed findings within the CSP–FP nexus have also been attributed to: methods, model specification and endogeneity issues (e.g. Li 2016; Orlitzky et al. 2003; Roberts & Whited 2013) and the social disclosure patterns across different industries and the institutional micro- and macro-settings (e.g. Garcia-Sanchez et al. 2016, Tshipa et al. 2018), due to associated costs and benefits with diverse industry features. The SA context makes an interesting case to examine voluntary social disclosures, specifically due to its long history of racial segregation, termed apartheid, which resulted in many social disadvantages to the majority of the SA population (Southall 2004). Consequently, Sampong, Boahene and Kweitsu (2018) contend that the practice of social disclosures has mainly been influenced by ‘apartheid’. Nevertheless, the unique CG codes have voluntarily promoted social disclosure in line with international standards. Specifically, the King III CG report recommends that all entities in SA should promote a voluntary disclosure regime based on the third-generation reporting standard (GRI, G3) (IoDSA, 2009). Supporters of social and global reporting standards have equally long contended that social disclosures offer reporting organisations a wide spectrum of intangible benefits, which extends to the firm’s balance sheet (Adams 2004). Although evidence of the CSP–FP nexus has generally been mixed, theoretical understanding generally supports a positive direction (Orlitzky et al. 2003). In the SA context, a study that examined inclusion on the Johannesburg Stock Exchange (JSE) socially responsible index and FP found no association using logistic regression (Du Toit & Lekoloane 2018). De Klerk and De Villiers (2012), using a modified Ohlson model, found that corporate responsibility reporting is significantly positively related to market value. Horn, De Klerk and De Villiers (2018) using survey data by KPMG International on CSR contend that social disclosure has a positive but weak effect on future profits. The advent of socially responsible behaviour, linked to international standardisation, through sustainability reports, ultimately seeks to ensure the long-term sustainability and survival of organisations, while maximising shareholder value and FP. Thus, socially responsible behaviour can be a competitive advantage to ensure firms’ long-term profitability. Socially responsible behaviour (and its related disclosure practices) is a management practice that ensures an organisation maximises the positive impact of its operations on society in a manner that meets and exceeds the legal, ethical and public expectations. Therefore, social responsibility creates a positive long-term value for shareholders even in conditions of uncertainty (McWilliams & Siegel 2000). For instance, Waddock and Graves (1997) found that an increase in CSR could lead to enhanced long-term profitability. Supporters of international disclosure standards on social responsibility have long contended that disclosure offers reporting companies a positive impact on FP. Studies specifically covering transparency and disclosure have indicated a positive market reactions, which could lead to a corresponding impact on firm value (Adams 2004). Therefore, based on the available empirical and theoretical evidence within the extant literature that socially responsible behaviour leads to a positive impact on firm profits (Orlitzky et al. 2003), we hypothesised that:

The call for appropriate representation of NEDs on boards (i.e. board independence) has largely been based on agency theory (AT) and RDT. Agency theory asserts that boards have a monitoring and controlling role over managers due to their opportunistic behaviour (Fama & Jensen 1983; Jensen & Meckling 1976). For instance, outside directors have more opportunity for control and face complexities stemming from their responsibilities as directors (Mangel & Singh 1993). Resource-dependence theory notes that the existence of boards serves as an essential link between firms and their critical resources needed to maximise value (Pfeffer 1973:350). For example, NEDs offer essential resources, by promoting board independence (Haniffa & Cooke 2002:319) towards voluntary social disclosures. Thus, RDT proposes that the unique composition of boards, including expertise, is a firm’s resource (Hillman & Dalziel 2003) that could affect its value (Pfeffer & Salancik 1978). In particular, RDT views the board composition as firms’ value-added resources. For instance, larger boards with professionals increase the firm’s expertise and resources (Pfeffer & Salancik 2003). Boards perform a critical supervisory and monitoring role, which is central to ensure that managers act in the best interest of shareholders (Fama & Jensen 1983; Jensen & Meckling 1976). In spite of the trade-off that may exist between social disclosure and FP, prior research argues that managers may undertake social responsibility activities for their own benefit, such as public fame, at the expense of their firms’ SWM goal. For instance, managers may seek to divert stakeholders’ attention from a poorly performing firm by improving its CSR activities (Preston & O’Bannon 1997). On the other hand, managers may reduce CSR overheads in order to boost short-term gains to enhance their personal rewards. These two views show that at the board management level, social disclosure could be used as a signal towards managerial benefits, which could have an adverse effect on firm value whenever managers act unfavourably towards stakeholder interest (Baird, Geylani & Roberts 2012). Therefore, towards re-aligning the principal agency conflicts is the promotion of good CG. Prior researchers have argued that CSR disclosure may have a limiting effect on firm value (Horn et al. 2018; Sampong et al. 2018) and that because of this possibility, the associated benefits of CSR disclosure for firms are largely dependent on good CG (Aguilera et al. 2008). Therefore, the existence of good CG, in particular BIND, could avert corporate irresponsibility. This could lead to an enhanced positive effect of voluntary SPD on FP. The reverse is a situation that can aggravate conflicts among various stakeholders (Hoje & Harjoto 2011) and could result in poor CDM benefits from ineffective CG. In the SA context, the promotion of BIND has been advocated by CG codes. For instance, formal CG regimes have sought to largely promote BIND. The 2009 CG code of SA, King III (IoDSA, 2009), requires boards to have the majority of NEDs to facilitate increased monitoring, towards shareholder and stakeholder CG-related disclosures (Gyapong & Afrifa 2019). The extant literature found a positive impact of the presence of NEDs on social disclosure and FP (Ntim & Soobaroyen 2013). From the available empirical findings, one can establish that the proportion of NEDs should potentially have a positive moderating impact on the association between social disclosure and FP. Arguably, the presence of NEDs on boards provides positive growth, with potentially positive implications for FP and voluntary disclosure (Ntim & Soobaroyen 2013). Hence, we hypothesised that:

Agency theory suggests that MO has the potential to align the divergent interests of managers with those of shareholders. Hence, extant literature empirical evidence shows that MO impacts on organisational outcomes. Johnson and Greening (1999), for instance, found a positive relationship between top management’s equity and CSP. The good management theory asserts that there exists a positive association between CSP and firm value (Graves & Waddock 1994); if so, then MO could promote a more active voluntary social disclosure among managers. Hence, the conclusion could hold that MO might enhance initiatives to engage more in voluntary social disclosure in support of FP. This is because managers have power to allocate resources among the wider stakeholder group; hence the provision of managerial incentives (e.g. stocks and bonds) could be an effective way to mitigate associated agency conflicts, and to help align the conflicting interest of agents to those of the owners (Fama & Jensen 1983; Jensen & Meckling 1976). More so, where managers own a significant part of the firm’s equity, decisions regarding promoting CSR could be better achieved (McConnell & Servaes 1990). However, it has arguably been noted that the different institutional contexts largely exhibited, due to wider differences in development (e.g. Garcia-Sanchez et al. 2016; Sampong et al. 2018), may affect voluntary disclosure differently in developing economies. Campbell (2007) argues that developed economies (like the United States and United Kingdom) may impose greater institutional pressure (Campbell 2007) on managers to undertake more socially related disclosures. However, in developing countries, like SA – where investor protection is relatively low as a result of a weaker external regulatory environment (LaPorta et al. 2000) – managers may not be subjected to mandatory forms of pressures. Unlike the United States and United Kingdom, ownership within SA companies is typically characterised by high block and institutional ownership, mainly in the form of complex cross-ownerships and tall pyramids. Nevertheless, SA CG codes encourage directors and managers to own shares in companies in order to improve interest alignment, including specifying performance-related elements of director pay (Barr, Gerson & Kanto 1995). For instance, evidence within extant literature supports a positive relationship between top management equity ownership and CSP (Johnson & Greening 1999). By implication, MO should enhance initiatives to engage more in voluntary social disclosures in support of long-term profits. This is because managers have power to allocate resources among wider stakeholder groups, and so, whenever managers own significant part of firms’ equities, decisions regarding promoting CSR and its related disclosure could be achieved much easier, which should lead to increased profitability. From the above, we hypothesised that:

The complementary theory (Aguilera et al. 2008:483) and CAP (Dierickx & Cool 1989; Galbreath 2016) both highlight the collective nature of boards and management in CDM. Complementary theory presents a more integrative view of CG that considers how governance elements interact and potentially complement each other (Aguilera et al. 2008) to impact on CDM, including voluntary disclosure. Inherently, extant literature notes that because there may be great difficulty and uncertainty involved in observing executive behaviours, as well as in measuring performance outcomes in corporations, managerial incentives and boards’ monitoring role may necessarily function as complements, as the two must be present for effective governance (Aguilera et al. 2008; Galbreath 2016; Milgrom & Roberts 1992). In a similar vein, and as implied by the RDT (Pfeffer & Salancik 1978, 2003), CG structures are critical firm resources that are used to achieve organisational goals. Hence for effective CDM, the firm’s resources must be uniquely combined to achieve success (Misangyi & Acharya 2014). This supports the CAP. The CAP suggests, for instance, that there are ‘strong relations of complementarity and co-specialization among individual resources, so that it is not really the individual resources but rather the way resources are clustered and how they interact’ (Foss 1998:143), which is important for competitive advantage. Thus, CAP holds that the effective applicability of the firm’s resources means that there exists a bilateral dependence among the resources (Galbreath 2016). As such, it has been argued that anything that is a strength or weakness to the firm can be considered as a resource (Wernerfelt 1984). For instance, Aguilera et al. (2008) suggested that ‘performance incentives for executives are more effective when complemented with a high level of board independence and an effective market for corporate control’ (as cited in Misangyi & Acharya 2014). Evidence on the effect of board independence on voluntary CSR and firm value supports a positive effect. Similarly, the impact of MO has been found to be positive. Arguably, these mechanisms are firm resources that aid the SWM goal. As such, it is believed that all individual units are contributors towards the firm’s SWM and its overall survival. Hence, we argue that the complementary interaction of the firm’s available resources is particularly necessary towards the achievement of optimal organisational gains. For instance, Aguilera et al. (2008) suggested that ‘performance incentives for executives are more effective when complemented with a high level of board independence and an effective market for corporate control’ (as cited in Misangyi & Acharya 2014). From the aforementioned theoretical and empirical evidence, we could conclude that CG mechanisms, from the perspective of utilising them as complements, should have a high marginal impact on the different organisational activities, compared with their individual effects. Thus, the board management monitoring roles and performance incentives must be present for effective governance and CDM (Galbreath 2016; Milgrom & Roberts 1992). Therefore, from the above, we could argue that the firm’s monitoring functions and incentive alignment are necessary to mitigate the agency conflict. Hence, the following hypothesis is formulated:

This study utilised an unbalanced data set of 126 listed JSE firms during the 2010 to 2015 period. Data were obtained from the African Market database. We excluded firms with more than 2 years missing data and all financial institutions, which produced total firm-year observations of 747. Our data set indicates both cross-sectional and time series components. Hence, given the nature of the data and to allow for time series and cross-sectional data observations, this study employs a panel regression technique. This is consistent with extant literature. South Africa is chosen as our study context because of its many unique features that seek to promote good CG and social disclosure. For instance, CG in the SA context has arguably been noted as well developed, which compares well with most developed and middle-income economies. Again, although the SA CG regimes stress the need to comply with affirmative actions, it equally endorses a voluntary disclosure regime based on international reporting standardisation (Sampong et al. 2018).

Table 2 presents a summary of descriptive statistics (i.e. mean and standard deviation) and the inter-construct Spearman correlation matrix for the study variables (excluding industry and year-dummies). It shows a mean of 1.974 for TQ, suggesting that, on average, most sample companies exhibited a positive long-term FP over the review period. Also, SPD has a mean of 0.690, suggesting that, on average, the extent of voluntary SPD among sample companies, on the basis of the GRI–G3.1 guidelines, was approximately 69%, which is reasonably high. Also, the mean for NEDs (BIND) is 0.272, suggesting that, on average, there are about 27% NEDs on the sample companies’ boards over the study period. Additionally, MO has a mean of 5.912, suggesting that, on average and across the sample firm-years, managers’ share-ownership was approximately 6%. Table 2 also presents the correlation among the variables. The Spearman correlation matrix was utilised as a measure of the strength and direction of association. Table 2 reveals the following associations: SPD is significantly negatively related to TQ. Non-executive directors is positively related to TQ and SPD, while MO is negatively related to TQ and SPD.

Our final data set produced unbalanced data as there were some missing years; hence, to take advantage of the variations in the variables of interest over time, an unbalanced regression panel is employed. However, before deciding on an appropriate estimation technique, and consistent with extant literature (e.g. Haniffa & Hudaib 2006), preliminary tests are performed. We assess multicollinearity using the correlation matrix (Table 1), which shows multicollinearity is non-existent. We also use the variance inflation factors (VIFs), which show a range of 1.02 to 0.98 (with tolerance levels ranging from 0.83 to 0.98). The range of VIFs falls outside the conventional threshold of 10 (Kennedy 1998). Lastly, we test for internal consistency of voluntary SPD index using the Cronbach’s alpha test. The outcome shows an overall coefficient alpha of 0.724, which is indicative of internal reliability (Potter & Levine-Donnerstein 1999). The unbalanced nature of our data set indicates that the use of an ordinary least square (OLS) regression method may not be appropriate, as it assigns equal weight to each observation and does not use information relating to unequal variability within the data. The generalised least square (GLS) regression, however, can produce more accurate estimators in the presence of outliers and unequal data variability, because it accounts for such information. It is equally able to capture firm-specific effects and also able to control for individual heterogeneity within the data set (e.g. Tshipa et al. 2018).

In Table 3 the model fit information shows that all the regression models are statistically significant which indicates a better model fit, and reflects the model’s adequacy. Wald’s chi-squared test is used to test the level of representation, which is also dependent on the significance levels, allowing for the acceptance or rejection of the models. Hence, all the models show significant adequacy. In Table 3, all models’ Wald’s chi-squared values are statistically significant (at 0.0001) for a 95% confidence level. This suggests that common method bias may not be a problem in our analysis (e.g. Podsakoff et al. 2003). In total, three models were estimated to test the stated hypotheses. However, consistent with prior studies (e.g. Galbreath 2016; Oh, Chang & Cheng 2018) we first test the control variables effects separately on TQ and SPD, in baseline Model 1. Our approach may slightly differ from previous studies, because this study examines the relationship between voluntary SPD and FP. The results are largely consistent with mixed findings within the extant literature: showing positive, negative and non-significant effects (see Table 3, Model 1 and Model 2).

This section attempts to deal with the problem of endogeneity, which has arguably been identified in the extant literature as the single most common issue in CG research. As such, ignorance of endogeneity issues may result in unreliable causality inferences (Roberts & Whited 2013; Tshipa et al. 2018). Arguably, variables are typically endogenous, instruments are scarce and causality relationships are complicated; nevertheless, it has equally been argued that most econometric methods utilised in extant literature are effective remedies to mitigate endogeneity issues (Li 2016). Therefore, we examine the robustness of our main findings in Table 3 compared to another modelling technique. Hence, we employ the seemingly unrelated regression (SURE) model, with a 1-year lagged independent variable. Thus, we introduced a 1-year lag between SPD and TQ, such that profitability in a given year (i.e. TQ_t) depends on the previous year’s voluntary SPD (i.e. SPD_t−1) (Sampong et al. 2018). Seemingly unrelated regression produces efficient estimates in cases of non-identical prediction equations. This is achieved by weighting the estimates by the covariance of the residuals from the individual regressions (Greene 2008; Kennedy 1998).

In Table 4, and consistent with the procedure followed in the main findings, three models were estimated to test the stated hypotheses using the SURE (with a 1-year lag of SPD) to test the robustness of our earlier findings, utilising the econometric model below: TQit=b0+b1SPDIit−1+b2SPDIit−1*NEDsit+b3SPDIit−1*MOit−1 +b4SPDIit*NEDsit*MOit+∑i=17b5CONTROLSit+εit[Eqn 3]

In Equation 3, b_0–5 are the intercepts, e is the error term, SPDI_it−1 is the 1-year lag of a main independent variable and CONTROLS are the control variables. All variables are as defined in Table 1.

The baseline Table 4, Model 1 tests the effect of all control variables on TQ and SPD producing R² values of 0.167 and 0.515. Table 4, Models 2–4 indicate a better model fit; particularly all chi-squared values are statistically significant at 0.0001. Table 4, Model 2 shows a positive (β = 0.078) direct voluntary SPD and FP association in the long term, with R² and ΔR² values of 0.170 and 0.003. Thus, again H1 is supported. Table 4, Model 3 reveals R² and ΔR² values of 0.187 and 0.018. Again, it shows that NEDs (BIND) positively (β = 0.108) moderates the direct positive voluntary SPD and FP nexus, when compared with the outcome in Table 4, Model 3 (β = 0.078). Again, the effect is non-significant. Table 4, Model 3 again shows a significantly positive (β = 0.393; p ≤ 0.01) moderating effect of MO on the direct positive voluntary SPD and FP nexus. Hence, H2 and H3 are all supported. Lastly, Table 4, Model 4 (with R² and ΔR² values of 0.194 and 0.006) shows that the interactive effect of NEDs and MO significantly positively (β = 0.398; p ≤ 0.05) moderates the direct voluntary SPD and FP nexus in the long term. The results in Table 3 and Table 4 show that the study’s findings are robust, with only changes in the coefficients in Table 4 due to the 1-year lag of the independent variable. Hence, the results are qualitatively robust.