Impact of CSR on Firm Value: The Moderating Role of Corporate Governance

This study aims to investigate whether the corporate governance (CG) moderates the link between corporate social responsibility (CSR) and firm value (FV). For this purpose, anatomization was conducted by extracting data from the published annual reports of non-financial firms listed on the Pakistan Stock Exchange. Correlation, regression, and moderation analyses were conducted to obtain the statistical outcomes. The results showed a significant direct relationship between CSR and firm performance. Additionally, it was found that the interactivity between CSR and FV weakened when CG was included as a moderator. The results of this study could be used by stakeholders to make economically sound decisions since it provides complete guidance regarding how to engage in productive CSR activities. Moreover, this study contributes to future research by examining the association between CSR and FV using CG as a moderator, in a market where, as in other developing markets, this relationship has not been the focus of research. Apart from its theoretical contributions, this study explores the role of CG as moderator, in line with research conducted in under-developed markets, which may be considered a significant contribution.


INTRODUCTION
In contemporary business environment corporate social responsibility (CSR) has become an imperative concern and debated a lot. Across the globe, numerous researchers are exploring its practical implementations in the various business scenarios. The debate on CSR has two aspects, opponent of CSR practices consider these activities as profit reducing practices for its shareholders and suggest that the core obligation of any business is to generate profits for its stakeholders by optimally utilizing firm's resources. On the contrary, the defender of CSR argue that its firm's responsibility to fulfill the needs of both investing and non-investing stakeholders who are directly affected by corporation's decisions and can also become a threat for the survival of business of African firms in the long run but negative in the short run. Firms initiating CSR activities go all-out to meet the expectation of all its stakeholders (Asmeri et al., 2017).
On the foundation of divergent corroboration suggested by past literature it is unambiguous that there is some kind of tie-up among CSR and firms' value although magnitude of that association is wavering concerning different studies. Therefore, based on these substantiations it is hypothesized that CSR has significant ability to govern the FV.
As maintained by Aebi et al. (2012) board independence has a significant moderating effect in the relationship between CSR and firms' performance. However, they further stated that board size has no significant effect on that relationship. Firms' value is significantly influenced by the CEO-Duality and it has a negative impact on dividend policy (Claessens & Yurtoglu, 2013). Some aspects of CG exert a significant impact on CSR disclosure (Odoemelam & Okafor, 2018).
Nonetheless, the effect of CG in firms' value and CSR relationship required more investigation for generalizability due to inconsistency between prior studies. Such as the study of Heremans (2007) investigates the relationship of firms value and cash holding in the financial institutions of UK, the finding of his study indicated the two dimensions of CG, board structure and board size has significant influence on that relationship whereas board meeting insignificantly moderating between firms' value and cash holding. Yet, in the literature, there are still gaps, which require more investigation to check their influence on different research models. Therefore, to fulfill this gap researchers took the overall CG dimensions between the relationship of CSR and firms value to investigate the moderating effect of CG in that relationship.
The purpose of this research study is to explore the impact of CSR on FV through moderating effect of CG by analyzing the published annual reports of listed non-financial firms at Pakistan Stock Exchange. Following are the research questions of the study: (1) What is the relation between CSR and FV? and (2) How CG mechanisms play moderating role among the interactivity of CSR and FV?
Researchers have scrutinized the association among CSR and FV including the CG as a moderator in a constituency where, as in several other developing markets, such sort of relation has n't been the focus of this research study. Apart from inclusion of CG as moderator, mainly keeping in view the research carried out in an under developed market, may be considered as a significant contribution. This study also unfolds the prominence of controlling the previously ignored variables in such sort of studies and spotlight the twist of mechanism through which the FV is affected by CSR ( Figure 1).

METHODS
To accomplish this research study, all non-financial firms that are listed at Pakistan Stock Exchange (PSX) are taken as the total population of the study. Total sectors listed on PSX are 35 from which only 7 sectors are financial and other 28 belong to non-financial sectors. There are a total of 573 companies that are listed on the PSX website dated November 23, 2018, from which 129 are financial firms and the remaining 444 belong to non-financial sectors of Pakistan.
From the above-mentioned population, 129 firms are excluded for the sample formation because these are financial firms and are not included in this study to analyze while the remaining 444 firms belong to the non-financial sector.
To construct the sample size, the authors collected data from 75 listed non-financial companies. Because of unavailability of data of some companies the sample size reduced to 71 companies. The study considers only non-financial firms because companies included in the financial sector have their own rules and regulations that are different from non-financial sectors (Firth et al., 2007).
Purposive sampling technique is used while selecting the sample size. In this technique, the investigator selected only those firms whose complete data were available in the form of published reports. The table is attached in the Appendix section consists of information regarding sectors that researchers used to make a sample for the current study.
For this study, researchers used data of 71 non-financial listed companies covering the time span 2008−2017. Ten years' data are collected from annual reports of non-financial listed firms to analyze the statistical outcomes.
Earlier studies suggest that CSR and firm performance get influenced by some other elements too, i.e., size and leverage of firm (Blazovich & Murphy Smith, 2011;Ullmann, 1985). These two variables are used as controlled variables in order to control their effect on FV. Tsoutsoura (2004) suggests size of the firm can be used as controlled variable since when firms increase in size many issues arise with it, including agency problems, monitoring and policy communication. Furthermore it is argued that huge sized firms are more concerned to invest in environmental and social cause and take initiatives for CSR practices more often as compared to small-and medium-sized firms (McWilliams & Siegel, 2001). Size is measured by taking natural log of total assets and this measurement is also endorsed by Tsoutsoura (2004).
Second controlled variable used in this is leverage, since it has indispensible influence on decision making liberty of firms. High leverage ratio may impose limitations on directors for investment decisions that may  (Barnett & Salomon, 2012). Furthermore, it is observed that mostly firms with lower debt ratio are more likely to perform CSR activities (Ishtiaq et al., 2017) (Table 1).
To draw statistical results and regression analyses following econometric equations are derived to accomplish the research objectives. At the very first level, study explains the relationship between CSR and firm performance. Second, the study analyzes the effect of CSR and firm performance while taking CG as a moderator. Following are the econometric equations derived for regression analysis to estimate the relationship among CSR, FP, and CG.
Where ROA: return of assets, i for time t Tobin's Q: market measure of firm, i for time t β°: intercept β 1 : intercept for each independent variable  by total assets. Tobin Q is the market measure to calculate firm performance and is calculated by market value of equity + book value of debt / book value of total assets. Table 2 explains the mean value of ROA for Pakistani firms which is 0.082 whereas the maximum value is 0.22 and minimum value is −0.066 with standard deviation of 0.078. Mean value of Tobin Q is 2.052 while the maximum value is 155.32 and minimum value is 0.04 with standard deviation of 8.09. The board size (BS) has a mean value of 9.151 while the maximum value is 19 and minimum value is 7 with standard deviation 2.314. B-IND having a mean value of 1.815 with maximum value 13 and minimum value 0 and standard deviation of 2.165.

RESULTS AND DISCUSSION
Family ownership (FMO) depicts the mean value at 2.059 however maximum value is 50.572 and minimum value is 0 with standard deviation 7.21. Largest shareholder ownership (L-own) has mean value of 37.793 while maximum value is 127.951 and minimum value is 0.37 whereas the value of standard deviation is 23.907. Donation (DON) has a mean value for Pakistani firms of 17,439.62, its maximum value is 149,237 and minimum value is 0 while the value of standard deviation is 26,524.04. The mean value of education (Hasan & Yun, 2017) is 0.804 while maximum value is 1 and minimum value is 0 and standard deviation is 0.397. Community development (CD) has mean value of 0.901 with maximum value 1 and minimum value is 0. The value of standard deviation is 0.298. The size of Pakistani firms has mean value of 23.704 while its range is from 29.467 to 19.595 with standard deviation of 1.435. Leverage of Pakistani firms has mean value of 0.168 while maximum value is 0.983 and minimum value is 0 whereas the value of standard deviation is 0.183. Table 3 reports the correlation coefficients among all variables of study and their p-values.
Controlled variable leverage has strong negative connectivity with size of the firm at 1% significance level. Dependent variable ROA indicating firms' performance is negatively correlated with size and leverage. It means firm performance and control variables, i.e., size and leverage of firm moves in opposite directions. Tobin Q is insignificantly correlated with leverage and ROA.
The board size is positively correlated with size of the firm and this bond is highly significant at 1% significance level. The board size has strong negative association with leverage at 5% significance level and ROA of the firm at 1% significance level while it has insignificant relation with market performance of the firm. Board independence has positive relation with size of firm and board size at 1% significance level while it has insignificant relation with leverage and market performance of firm. At the same time independent directors in board has negative correlation with ROA and this correlation is significant at 10%.
Family ownership has insignificant tie-up with size of firm and market performance indicator Tobin's Q. FMO has strong positive correlation with controlled variable leverage of firm at 1% significance level and has  182*** 0.179*** 0.534*** 1.000 Notes: *** p < 0.01 represents 1% significance level, ** p < 0.05 represents 5% significance level whereas *p < 0.1 represents 10% significance level negative correlation with accounting measure of firm performance ROA which means performance of family-owned firms is lower than non-family-owned firms. Table 3 further reports that FMO has positive correlation with board size and independent board of directors.
Largest shareholder ownership is positively engaged with size of the firms and is negatively correlated with leverage. While L-own is strongly and positively correlated with ROA & TQ, indicating largest shareholders have direct impact on the performance of firm. BS and independent board of directors both have insignificant correlation with L-own. The table also indicates that L-own has strong negative correlation with FMO.
Donations have significantly positive association with size and leverage of firm whereas it has insignificant correlation with ROA, TQ, and BS. DON is positively correlated with independent board of directors' means non-executive directors in firm are more concerned toward CSR activities and social responsibilities of the firm. Donations are negatively correlated with FMO and L-own. The association among education and the size of the firm is of direct nature and has insignificant association with leverage. EDU is positively engaged with ROA and independent board of directors', however, education shows insignificant relation with TQ, BS and FMO. Community development shows direct sort of interactivity with size, ROA, L-own and independence of board. CD has indirect association with leverage while it is insignificant interactivity with TQ and FMO.
To understand whether CSR activities increase the performance of firm or not linear regression tests are performed on the predicting variables. Performance of firm is measured through ROA and Tobin Q.
Model 1 of linear regression table reports the direct impact of CSR mechanisms on dependent variable FV. First variable of CSR is donations. Its coefficient is positively correlated with ROA and this association among both variables is highly significant as its p value is 0.001. This indicates that CSR activities have their direct impact on FV in such a way more a firm involves in CSR initiatives, more the performance of firm enhanced. This particular result designates that when firms take initiatives to engage in CSR activities and use their resources for the betterment of society, it will help the firm to maintain good relations with community and hence this goodwill helps the firm to improve its performance. The outcome of study is supported by previous studies as well (Ehsan & Kaleem, 2012;Makni et al., 2009;Orlitzky et al., 2003;Preston & O'bannon, 1997).
The dummy variables for CSR measurement used in the study are education and community development, both show significant coefficients with ROA accounting measure of firm performance. Education is significant at 1% while community development shows significant result at 10% with ROA. Hence, the studies by Choi et al. (2010), Fu and Jia (2012)) also confirm that firms who are socially active and invest in socially responsible activities enjoy the edge of good firm performance in return from society.
However, controlled variable size has an inverse and significant relationship with ROA as its coefficient is (−0.011) at p-value (0.000) which means in large-sized firms management fails to monitor all the issues whose effect is firm performance. The result is consistent with a previous study that increased firm size and is negatively related with FP (Donker et al., 2008).
The leverage being second controlled variable is negatively associated with ROA with coefficient value (−0.186). The negative outcome indicates that firms with good financial performance mostly use less amount of debt for the operations of firm. The above findings are confirmed by Demsetz and Villalonga (2001), Himmelberg et al. (1999) and Welch (2003) as they stated leverage and FP are inversely related to each other.
The value of R-square is 0.193, which means that 19.3% variation in firm performance (ROA) is explained by CSR while remaining variation is because of unnoticed elements. However, results also designate that model is a good fit because the value of Prob > F is 0.000.
Model 2 of simple regression in Table 4 highlights the results of CSR impact on Tobin's Q (market indicator of firm performance). The findings indicate that direct impact of CSR on FP (Tobin's Q) is positive and significant with p-value (0.000). Both dummy variables education and community development for CSR measurement also have positive coefficient values 0.403 and 0.196, respectively. Education shows significant results with FP at 1% significance level whereas community development shows insignificant result with p-value 0.168 but overall it can be concluded that CSR has significant positive impact on FP. These results are consistent with the findings of McGuire et al. (1988), McWilliams and Siegel (2001), Waddock and Graves (1997). When firms invest for social cause their market value is also enhanced as they are considered as socially responsible firms thus the enhanced image ultimately uplifts the financial performance of the firm. These findings also support the findings of Jiao (2010) and Jo and Harjoto (2011).
The size and leverage again have inverse association with FP (Tobin's Q) at 1% significance level. Whereas the value of coefficient of determination (R-square) is 0.053, which indicates that only 5% variation in firm performance is because of CSR. Whereas the remaining 95% variation is because of those variables that are not included in the study. The value of F test point out that the model is a good fit (Table 5).
Model 3 initially explains the direct effect of moderating variable CG mechanisms on dependent variable ROA then it describes the indirect effect of CSR on ROA with moderating role of CG. The total directors in the table means the total number of directors in board indicates board size. So, we take total directors as "Board Size" while explaining the results.
The table reports the direct impact of board size on ROA and Tobin's and suggests that BS has inverse impact on firm performance with p-value (0.000) indicates strong significant results. It means that when the board size is increased in size, decisions regarding different policies become difficult to take and implement due to difference of opinion and agency problems arise. Furthermore, policies and decisions are delayed due to huge board size thus overall effecting the financial position of the firm. This negative association among board size and firm performance is consistent with the finding of Ujunwa (2012). "Donations" have insignificant relation with ROA. This association among both variables is changed from significant to insignificant because of the presence of moderating variable "Board Size." The results of the table also report inverse effect of board independence on firm performance (ROA and Tobin's) at 1% significance level, it indicates that independent boards in Pakistani firms are not able to improve firm's performance because of their weak monitoring policies and low engagement in firm's operations. The findings are consistent with results of Garg (2007). The impact of donations on ROA is insignificant in the presence of board independence but its impact on Tobin's Q is significant at 5% significant level in the presence of the moderator, thus CSR does not change its direction on market indicator of firm performance but it does The results of the table represent that family ownership has direct linkage with firm performance with coefficient values 0.001 and 0.002 for ROA and TQ, respectively, but this association is insignificant in both cases. This result is consistent with the study by Villalonga and Amit (2006). The interaction term shows 10% significant result for ROA and insignificant result for TQ with positive coefficients. It means that CG and CSR combined have little effect on the performance of the firm. While in the presence of FMO, CSR has positive significant effect on FV. Thus, findings suggest that firms with good family composition in higher management are able to perform better in financial terms.
Direct effect of L-own on ROA and Tobin's Q is strongly significant, and it is confirming the facts that ownership concentration enhances firm performance. When interests of executive management and shareholders are aligned and there is no conflict of interest, this situation reduces agency costs and increases the FV at larger. The result is consistent with previous studies as well (Morck et al., 1988). Findings of studies in emerging economy also supports the result of current study that largest shareholding positively enhance company's performance (Yasser & Mamun, 2017). However, interaction term between largest shareholder's ownership and CSR is also positively significant. Thus, results suggest that intervening term significantly and positively impact FV, which means the existence of larger shareholders encourage the top management to invest more in society concern activities and thus it increases the performance of the firm. The result is consistent with Ishtiaq et al. (2017).
In model 4 the authors analyzed the effect of moderating variable CG with CSR dummy variable "Education." Here, board independence shows positive significant relation with ROA while it has insignificant relation with Tobin's Q. Significant result indicates that B-IND enables non-executive directors to perform their unbiased role in monitoring the firm's operations and hence improve firm performance. This positive result is consistent with previous studies (Bhagat & Black, 1999;Mashayekhi & Bazaz, 2008;O'connell & Cramer, 2010). Whereas interaction term (B-IND and Education) has positive significant effect on ROA at 1% significance level which means that when non-executive directors play their role efficiently while accompanying for societal activities it boosts the performance of firm as a whole. While the moderation relation of B-IND and CSR dummy variable Education does not prove its relation with market performance Tobin's Q as a result shows insignificant relation.
The table explains the impact of BS on ROA and Tobin's Q in moderation regression analysis, its relation becomes insignificant with FP. The value of coefficient is still negatively associated with ROA but significance of relationship is changed. Furthermore, dummy variable Education also becomes insignificant with ROA in the presence of BS and their interaction term also show insignificant result which means that overall the moderation of BS is not proved in creating a link between CSR engagement and FP. While dummy variable Education is showed strong significant and positive relation with TQ in the presence of BS. This result is consistent with resource dependency theory as well as that larger boards have diverse skills based on their experiences to improve financial position of firm and they play key role in efficient monitoring of firms. Thus increased BS helps the firms to grow financially as they have external connections and access to capital market can reduce the uncertainties for firms (Goodstein et al., 1994;Haniffa & Hudaib, 2006). The interaction terms for BS and education has negative significant result with TQ. While the direct impact of education on ROA is positive but now because of the interaction of moderator term between education and FP, their strength and direction of relationship is changed from 1% to 5% and from positive to negative association.
The table reports that FMO has positive but insignificant relation with ROA and negative but insignificant relation with TQ. Most of the studies indicate a strong family-owned culture in Pakistani firms and most of the 156 Butt et al. businesses are monitored by family members (Yasser et al., 2011). While on the other hand, Villalonga and Amit (2006) debated that when FMO exceeds from a maximum limit then it becomes the reason to drop shareholders value and thus effecting the performance of the firm. Therefore, we can conclude that family-owned firms have contrary views to describe their association with FP. Direct effect of education (a dummy variable of CSR measurement) has strong positive effect on ROA and Tobin's Q but their association with education in the presence of FMO is positively insignificant which means that moderation of family ownership as a mechanism of CG is not proven in creating a linkage between CSR and FP from the results of this study.
The table reports the impact of large shareholders on FP through ROA and TQ. The results of study indicate that the largest shareholder has positive insignificant relation with ROA and negative insignificant relation with TQ while the effect of education is also not significant. Thus, the impact of education on ROA and Tobin's Q is also insignificant in the presence of moderating variable largest shareholder ownership. This insignificant relation is consistent with the studies by Ishtiaq et al. (2017).
Controlled variables "size and leverage" have negative correlation coefficient with ROA and Tobin's Q. The value of R-square indicates that 20% variations in ROA is because of CSR and 8% is because of Tobin's Q. The value of Prob > F shows that the model is a good fit.
Model 5 indicates the results of BS on ROA and Tobin's Q where BS has insignificant relation with both firm performance and market returns as well in moderation regression analysis. This result is changed because of moderating factor in analysis. Whereas community development depicting significant relation with both ROA and Tobin's Q. This relation is consistent with previous study that corporations who invest more in community spheres are more likely to have increased firm performance (Ishtiaq et al., 2017). The interaction term explains at 10% significant value negative role of BS with coefficient value of (−0.008) with ROA whereas it shows at 10% significant value negative results with coefficient value (−0.133) with Tobin's Q in explaining the impact of CSR on FP. Hence, the negative value indicates that in the presence of BS as a moderator the direction of relation among CSR and FP is changed from positive to negative which means that now CSR engagement decreases the firm performance when BS is taken as a moderator.
The result of B-IND shows insignificant results for both ROA and TQ with positive and negative coefficient values, respectively. While community development also shows insignificant result with firm performance with coefficient values 0.036 and 0.256. This insignificant result is because of the presence of B-IND as a moderating variable in the model. While interaction term results indicate that combined effect of community development and B-IND is negative and insignificant with ROA and TQ at coefficient values −0.017 and −0.0256. The result becomes insignificant because of moderation of B-IND. Hence, we can say that B-IND does n't play a significant role in creating a link between community development concerns and firm's performance.
The table reports that ownership structure has no significant effect on FP. FMO has negative results with coefficient values −0.001 and −0.019 with FP in both accounting measure (ROA) and market measure (Tobin's Q). However, community development also shows insignificant results with FP measures in moderation regression analysis. Further results of interaction term are also insignificant with coefficient values 0.001 and 0.010 with ROA and TQ. Thus, it means in analyzing the effect of FMO as a moderator the combined effect of family ownership and community development is insignificant thus proving no significant role of ownership structure in analyzing the impact of CSR and FP.
The results of largest shareholder's ownership with effect of community development on ROA and Tobin's Q are reported in this table. According to the results of the table, ownership variable L-own has negative insignificant effect on ROA with coefficient value (−0.001) and it has positive significant relation with Tobin's Q with coefficient value (0.006). Effect of community development on ROA and Tobin's Q is also insignificant. Khan et al. (2013) also shows insignificant relation between ownership concentration and FV. Furthermore, the results of interaction term are also insignificant where the authors analyzed the combined effect of moderating variable L-own and independent variable CSR's dummy measure community development. Thus, concluding that moderation of largest shareholders in defining the effect of CSR on FP is not proved through moderation regression analysis.

Studies of
Controlled variable "size" and "leverage" are negatively correlated with ROA and TQ thus indicating that larger firms and firms with larger amounts of debts both are unable to improve the firm performance. The value of R-square explains 19% variation in ROA is because of CSR and 6.5% variations in TQ is because of CSR engagement. The value of Prob > F shows model is a good fit.

CONCLUSION
This study attempted to maintain a CSR framework for quantifying the CSR practices while using CG as a moderator. The theoretical contribution of this research is ensuring the support for signaling theory. To assemble the data for this study, the secondary data source is used such as published annual reports. Moderation regression analysis were used to determine the statistical outcomes which is also a contribution of this study in the field of finance as literature contributes much less work on moderation analysis in an emerging economy such as Pakistan.
The current study focused on different elements of CSR like donations, education, and community development for examining its impact on FV. By composing the results of the three above-mentioned elements of CSR it is confirmed that they have a great effect on firm's performance, i.e., enhanced firm performance. This study identifies the role of different attributes of CG as a moderator. The results show that CG has not played a vital role in elaborating the relationship between CSR and FV. It means that role employed by CSR on FV can't be temperate by CG. The results of the study are consistent with previous studies in analyzing a strong impact of CSR on FP (Ishtiaq et al., 2017).
Most of the previous studies have scrutinized CSR and its by-products only in well-developed countries. In this study the authors examined the association between CSR and FV in relatively less developed countries. It has been examined that the interactivity among CSR and FV including CG as a moderator in a constituency where, as in several other developing markets, such sort of relation has n't been the focus of research. Apart from inclusion of CG as a moderator, mainly keeping in view research carried out in an under developed market, may be considered as a significant contribution.
The outcomes of this exploration could be used by all stakeholders to make economically fitting decisions since it provides an absolute guidance about how to engage in productive CSR activities. This study also unfolds the prominence of controlling the previously ignored variables in these studies and spotlight the twist of mechanism through which the FV is affected by CSR. This research focused on the extent of CSR at PSX in 71 non-financial companies only and this study is limited to measure the firm performance through ROA and Tobin's Q only. Further studies on comparative analysis between different sectors like manufacturing and service sectors can be carried out. Many other measures of FP such as ROE, EPS, and NPM, can be used to measure the firm performance.