Apart from other corporate decisions, shareholders are primarily concerned with dividend policy decisions, which are a central concern in firms’ financial decisions of the firms. ²⁴ It is evident from the existing literature that, in firms’ corporate governance, dividend policy has received greater importance as an influential key ²⁵ and in Chinese initial public offerings (IPOs), CG is an important determinant of dividend policy. ^{26, 27} From the agency theory perspective, in firms with higher ownership concentration, minority shareholders are expected to be at expropriation risk when large shareholders, by using their voting rights, tend to affect the firm’s policies and decision-making for their private benefits. The threats of expropriation are more common in environments (e.g., emerging economies) with weak CG and low investor protection according to Ref. 3. Moreover, large shareholders prefer to generate their own benefits when they gain full control of the firm and do not share these benefits with minority shareholders. ²⁸ In East Asian and European economies, prominent agency problems arise due to minority shareholders’ expropriation by controlling shareholders. ²⁹ The dividend policy is significantly determined by agency costs and can be used to limit minority shareholders’ expropriations by reducing corporate wealth under insiders’ control. ^{30, 31}

The ultimate goal of CG is to ensure that investors who supply finances to firms receive their investment returns is the ultimate goal of CG. ²⁸ Shareholders receive return on investment either in the form of capital gain or dividends however, they prefer dividends to overcome the expropriation fear from controlling shareholders, as reported by Ref. 32 that, “dividends (a bird in hand) are better than retained earnings (a bird in the bush) because the latter might never materialize as future dividends.” Our discussion of the link between CG and dividend policy is based on the two most widely used agency models of dividends proposed by Ref. 32. The first model known as “the outcome model” of agency theory suggests that to restrict managers’ private benefits of free cash flows, firms with strong CG tend to pay higher dividends. In other words, shareholders have the power to influence dividend policies when they have greater rights through strong CG or legal protection. ¹⁰ Hence, better governance helps firms reduce agency conflicts through higher dividends, meet shareholder demands, and enhance their trust in management, and consequently, the firm. The outcome model predicts a positive association between dividend policies and CG quality. Many studies have supported this model. For example, using a sample of Canadian firms, Ref. 33 showed that dividend yield is positively associated with board composition and governance quality. Reference 34 report that in the case of US and Indian companies, dividend payouts are positively associated with governance quality. They analyzed the interrelation between shareholder protection at both firm and country levels. Reference 35 claims high payouts in strong CG firms for a sample based on emerging markets. Based on survey data from listed firms in Korea, a positive association is found between good CG and dividend payouts. ³⁶ Reference 32 find support for the outcome model by using a sample of 33 countries and report that in countries with strong investor protection, firms have a higher dividend payout ratio than countries with poor investor protection. Dividend payouts are higher and more consistent in firms where CG is stronger. ^{25, 37} Reference 38 assert that firms with stronger shareholder rights pay more dividends because dividend policy is expected to be the outcome of the governance regime. Reference 39 found a positive association between dividend payouts and strong shareholders in Dutch firms. In countries with stronger creditors’ rights, firms tend to have higher payouts ⁴⁰ supporting the outcome model. Recently, Ref. 41 studied the influence of family engagement and CG on dividend policies. They explore non-financial companies in Morocco and reveal that GC mechanisms have a strong influence on divided policy in Morocco.

Conversely, the substitution hypothesis of agency theory proposes that, to have capital market excess and maintain good shareholder relations, higher dividend payouts can be used as a substitute for weak governance. ^{42, 43} In firms with weak shareholder rights, building reputation by establishing good relationships with investors is a key concern, and dividend payouts play a vital role in building this reputation. By contrast, reputation and dividend payouts are less needed in firms with strong corporate governance. ⁴⁴ Therefore, firms with weak CG should have large dividend payouts to compensate for their weak shareholder rights. Since dividend payouts help to reduce managerial expropriations through a decrease in free cash flows, dividend payouts and CG are expected to have an inverse relationship ³⁸ and Under weak governance, dividends act as substitutes for poor investor protection. Many studies have demonstrated the substitution effect of dividends. For example, CG strength is inversely associated with dividend payouts. ^{38, 45– 48} The strength of corporate governance is measured by using CG index of Gompers (2003). ⁴⁹ Reference 50 analyzed a sample of master limited partnerships and reported that low-quality governance firms make high dividend payments; however, these payments are value-destroying, as they negatively affect firms’ cash holdings and firm value. CG can be effectively substituted by dividends to reduce agency costs. ⁴⁸ Reference 51 found a significant negative association between dividend policy and CG mechanisms, such as being familiar with auditors, timeliness, and shareholder rights. Managerial entrenchment is positively associated with both the level and propensity for dividend payouts. ⁵² Using the governance index of Gompers ⁴⁹ along with institutional block holdings and board structure, Ref. 45 report a negative relationship between dividend payouts and governance quality in favor of the substitution effect. In contrast to low-quality governance firms, firms with high quality governance predicted as dividend payers tend to make fewer dividend payments. ^{47, 53} He further argued that, in firms with weak governance, dividend initiation receives a positive response from shareholders, indicating the substitution of dividends for weak governance.

In Chinese firms, prominent conflicts of interest exist between minority shareholders and large dominant shareholders, and there are more expropriation opportunities for majority shareholders due to weak investor protection and creditor rights. ^{6, 54} Minority shareholders are more likely to suffer from self-serving behavior, such as excessive compensation, expropriation, and tunneling. Because Chinese firms suffer from weak shareholder protection and poor CG (e.g., minority-majority shareholder conflicts and state-owned equity control), they tend to pay high dividends as a governance control mechanism. ^{4, 6} In other words, Chinese firms are likely to use dividends as substitutes for poor governance to reduce conflicts between minority and majority shareholders. Considering Chinese firms as a sample for our research study, we develop the following hypothesis: H1:

Firms with strong (weak) corporate governance tend to pay low (high) dividends.

The growing interest in corporate social responsibility and increased importance of dividend policy in corporate finance literature based on its “puzzle” like nature have attracted researchers’ attention to explore the nature of link between them and the theories underlying this relationship. However, most of the literature (e.g., Refs. 14, 55, 56, 57) focuses on developed economies, and only a few studies have explored this relationship in emerging economies, such as China and few others (eg., Refs. 58, 59). According to Ref. 56, the link between CSR and dividend policy is guided through two channels: the earnings channel and equity cost of the capital channel. Our hypotheses on the CSR dividend policy link are based on these two CSR views on dividends.

According to the cost of equity capital channels, instead of paying dividends, firms tend to hold or invest cash because their involvement in CSR activities can reduce the equity cost of capital through its risk premium. ^{56, 60} A firm’s CSR practices lower its financial constraints either by reducing the cost of bank loans or the cost of capital. ^{61, 62} High CSR reduces equity capital costs by lowering the associated risk premium because socially responsible firms tend to enjoy good reputations, ^{63, 64} have good management practices and low information asymmetry, ^{57, 65} align interests between shareholders and other stakeholders, ⁶⁶ loyal customers or investors ^{67, 68} and have low risk and good governance. ⁶⁹ Since the opportunity cost of holding cash is low (high) in the case of a low (high) equity cost of capital, firms’ incentives to hold cash increase (decrease). ⁷⁰ On the other hand, firms might have more (less) incentives to use cash for investment purposes because financial constraints are less (more) strict for firms in the case of lower (higher) equity cost of capital. ^{56, 58, 71} Moreover, being financially less constrained, CSR firms can invest in projects that appear unprofitable and unfeasible because of high capital costs and low availability of external finance. These arguments suggest that CSR involvement reduces the equity cost of capital because firms are more likely to hoard or invest cash instead of paying dividends. In other words, the equity cost of the capital view of CSR and dividend policy links predicts a negative association between CSR and dividend policy.

In contrast to the first CSR view of dividends, the earnings channel view argues that CSR affects dividend policies through its effect on earnings. Earnings have been identified as a crucial determinant of dividend policy and firms with greater earnings ability are more likely to pay dividends. ^{72, 73} Socially responsible firms tend to have enhanced earnings, which leads to higher dividend payments. The mechanisms through which CSR-related activities enhance firms’ earnings include efficient management and good relationships with stakeholders. ^{74– 76} A better firms-stakeholder connection helps in developing competitive advantage, ^{77, 78} reducing transaction costs, ⁷⁹ cash flow shocks in the occurrence of some negative events ⁸⁰ and lower firm risk. ⁵⁶ In addition, CSR activities are associated with higher ethical standards and greater transparency, ^{81– 83} enhanced corporate reputation, ⁸⁴ greater customer loyalty, high commitment of employees, a high degree of suppliers’ cooperation ⁸⁵ and a reduced annual labor turnover rate. ^{86, 87} These factors significantly contribute to the earnings-generating ability of firms with strong involvement in CSR activities, which, in turn, increases their dividend payouts and enables them to pay dividends (if non-payers). Based on the preceding arguments, firms’ involvement in CSR activities can result in dividend payouts through increased earnings, as CSR activities are considered positive net present value projects. Hence, stronger CSR involvement puts firms in a better position to pay dividends and to pay higher dividends. The earnings view predicts a positive association between CSR and dividend policy. Reference 88 examined CSR and dividend policies in India. They revealed that CSR has a constructive influence on dividend disbursements. In addition, ⁸⁹ explore how CG influences the link between information asymmetry and dividend disbursement in an emerging economy. They found that dividend payouts are inversely inclined by information asymmetry issues. They also find that the link between information asymmetry and dividend strategy is less prominent in companies with robust CG systems. This finding is in line with the argument that companies encounter lower agency and asymmetric information issues when they pay higher dividends.

From the perspective of Chinese firms, evidence on the link between CSR and dividend payouts remains ambiguous. For example, Ref. 58 provide evidence in favor of earnings channels, while Ref. 12 support the equity cost of capital channels. Therefore, based on the two divergent views on the link between CSR and dividends, we develop the following hypotheses: H2a:

Firms with involvements in CSR activities tend to pay low dividends (equity cost of capital channel).

As the function of CSR is to make firms accountable to their stakeholders, firms committed to CSR practices can identify and assess the demands of different stakeholders. These CSR commitments facilitate a strong link between a firm and its stakeholders by considering their needs and demands. According to Ref. 90, CG and CSR are interdependent, and CSR as a firm’s obligation to society can be regarded as an external mechanism of corporate governance. The main argument in this paper is that CG is likely to influence the corporate social responsibility (negatively/positively), which in turn influences dividend payments. Our expectation of CSR’s mediating role in the link between CG and dividend policy is based on two arguments. First, in the context of stakeholder theory, firms with effective CG invest more in CSR to ensure the protection of all stakeholders beyond shareholders. ^{91, 92} Further, the earnings channel predicts that firms with stronger CSR involvement are expected to pay higher dividends. ^{83, 93} Based on these theories, it is argued that better CG can result in high CSR involvement, which in turn leads to high dividend payments through the increased reputation and earnings of the firm. Therefore, CSR is expected to facilitate a link between CG and dividend policies.

Second, stakeholder theory suggests that better-governed firms provide interest protection to all stakeholders beyond shareholders by investing more in CSR. In addition, the cost of equity channels suggests that firms with stronger CSR involvement tend to pay low dividends, as in this case, firms prefer to hoard or invest cash instead of paying dividends due to the low cost of equity. ^{56, 94} Accordingly, it is argued that better CG can result in high CSR involvement, which, in turn, reduces dividend payments because of a firm’s preference to hoard cash or invest it instead of dividend payments. In this context, we expect CSR to mediate the link between CG and dividend policy. The following hypothesis was developed to test our arguments regarding mediation: H3:

CSR mediates the link between corporate governance and dividend policy.

Based on the arguments above, the conceptual framework of mediation to explore the link between CG and dividend policy with the mediating impact of CSR is shown in Figure 1.

Firms listed on the Shanghai and Shenzhen stock exchanges are considered as a study sample to investigate the link between CG and dividend policy with the mediating role of CSR. The sample is selected using the 2012 China Securities Regulatory Commission (CSRC) industry classification, under which the firms are categorized into 19 different industrial sectors. The data collection period ranged from 2012 to 2021 based on data availability and an attempt to use the most recent data as much as possible. The final sample included 15,800 firm-year observations.

CSR data were collected from the Hexun database, which offers the CSR evaluation score of listed Chinese firms beginning in 2010. Many previous studies have used CSR data from this database. ^{58, 95– 98} To appraise the corporate social responsibility performance of Chinese firms, Hexun.com ranks first, and provides a good and comprehensive source of CSR data. ^{58, 96} Moreover, it comprehensively covers all critical stakeholder groups, including shareholders, employees, suppliers, customers, environment, and community in its assessment framework. ⁹⁶ The data on all other variables, including CG and control variables, were collected from the China Stock Market and Accounting Research (CSMAR) database.

The dependent variable is the dividend policy, for which the dividend payout ratio (DPR) is used as a proxy. Many researchers have used the dividend payout ratio as a proxy for dividend policy for example, Refs. 36, 56, 58 and 99– 102. The dividend payout ratio (DPR) was calculated as dividends per share divided by earnings per share. We use this measure because, according to Refs. 24 and 93, it establishes a direct connection between distribution and earnings.

The independent variable corporate governance was measured using the corporate governance index (CGI). CG is considered a multidimensional notion, defined as the mechanism used for the protection of different stakeholders’ interests. ¹⁰³ Many cultural, economic, environmental, legal, historical, and political factors affect corporate governance. This study builds the CGI using 35 items adapted from the prior literature of Asiedu and Mensah (2023), ¹⁰⁴ Ledi and Ameza-Xemalordzo (2023), ¹⁰⁵ and Munisi and Randøy (2013). ¹⁰⁶ The items cover key aspects of corporate governance, including board structure, audit and remuneration committees, transparency and disclosure, and shareholders’ rights. To calculate the CGI, we first computed sub-indices for each governance dimension. Each item was assigned a score of 1 if the statement applied to the firm and 0 if it did not. The sub-indices A detailed list of items along with coding rules is provided in Table 1.

Corporate Social Responsibilty (CSR) is used as a mediator. This is calculated using firms’ CSR scores from the Hexun database. This database calculates the CSR scores of Chinese listed firms based on five dimensions: employees, environment, community, supplier and customer, and investors. Each of these dimensions further consists of different sub-dimensions and is given different weights depending on the industries to which the firm belongs. Consistent with previous studies (eg. Refs. 58, 96, 98, 107), we calculated CSR scores as a total of four sub-social responsibility scores: employees, environment, community, supplier, and customer. First, the individual score for each sub-dimension was calculated, and then the aggregate score was calculated as the average of the sub-dimensions’ individual scores.

In addition, we control for previously identified factors that can affect dividend policy. To consider the firm’s financial condition, we include firm size, leverage, market-to-book ratio, profitability, ratio of retained earnings to total assets, and research and development expenditure as control variables. Firm size (SZ) is expected to affect both firms’ debt and dividend policies. ¹⁰⁸ Larger firms are more diversified, have less volatility and more regular cash flows, and can maintain debt at high levels, as they are less likely to go bankrupt ¹⁰⁹ and large firms with good financial conditions usually pay dividends. ¹¹⁰ Because leveraged firms are riskier in their cash flows and try to avoid external capital costs ^{16, 111} therefore firms’ dividend behavior is also expected to be affected by leverage. According to Ref. 112, through the market to book ratio (MTB), firms convey information to shareholders about their smooth dividend payments; hence, MTB is also added as a control variable. Moreover, profitability (ROA) is also considered an important determinant of dividend policy in the literature and is expected to influence dividend behavior because it reflects a firm’s resource availability after investments are funded. ^{6, 113} R& D expenditure is used as a proxy for firm growth opportunity and is expected to influence dividend policy because as compared to non-growth firms, growth firms pay fewer dividends. ^{114, 115} Further, the ratio of retained earnings to total assets (RETA) significantly determines dividend payouts, and firms with a higher proportion of earned capital instead of contributed capital tend to pay more dividends; therefore, Ref. 116 added RETA as a control variable in the model. A summary of the variables, their calculations, and data sources is provided in Table 2.

To investigate the mediating impact of CSR on the causal link between CG and dividend policy, we used a fixed-effects regression model identified through panel data model identification tests. Our analysis is based on the mediation approach suggested by Baron and Kenny, ¹²⁰ which consists of the following steps. DPR it = β 0 + β 1 CGI it + β 2 SZ it + β 3 ROA it + β 4 MTB it + β 5 LVG it + β 6 RETA it + β 7 R & D it + ε it (1) CSR it = β 0 + β 1 CGI it + β 2 SZ it + β 3 ROA it + β 4 MTB it + β 5 LVG it + β 6 RETA it + β 7 R & D it + ε it (2) DPR it = β 0 + β 1 CSR it + β 2 SZ it + β 3 ROA it + β 4 MTB it + β 5 LVG it + β 6 RETA it + β 7 R & D it + ε it (3) DPR it = β 0 + β 1 CGI it + β 2 CSR it + β 3 SZ it + β 4 ROA it + β 5 MTB it + β 6 LVG it + β 7 RETA it + β 8 R & D it + ε it (4)

First, the effect of the independent variable (CGI) was tested on the dependent variable (DPR). Here, the independent variable was expected to show a significant relationship with the dependent variable. In the second step, the link between the mediation variable (CSR) and the independent variable is tested, and the independent variable is expected to have a significant influence on the mediator. In the third step, the effect of the mediation variable on the dependent variable was tested, and the results needed to be significant. In the final step, both the independent variable and mediator were tested on the dependent variable. Here, the effect of the independent variable (beta coefficient of CGI in the final regression model) on the dependent variable is expected to be less in absolute terms compared to the effect in the first regression model. ¹

Table 3 presents the summary statistics of all the study variables. Summary statistics provide an overall picture of the data, including the mean, standard deviation, and minimum and maximum values of the variables for the sampled Chinese firms. The dependent variable DPR has a mean value of 0.227, which means that Chinese firms pay, on average, 22.7% of their earnings in dividends. The average value of DPR is closer to 0.268 (26.8%), as documented by Ref. 58 for a sample of Chinese listed firms. Although this value of the payout ratio falls within the range (20%-30%) fixed by the CSRC, it is quite low. The reason behind This is because minority shareholders’ rights in China are not strong enough to pressure firms to pay dividends. The average CGI score is 0.558, ranging from a low score of 0.221 to a high score of 0.814, which means that, on average, sample firms meet half of the governance standards. The average CSR score of Chinese listed firms is 25.68, which is quite low. The mean value of firm size is 22.118 (similar to 21.60 reported by Ref. 121 and ROA is 0.052, which means, on average, our sampled firms are relatively large and less profitable. The sampled firms have an average debt ratio of 44.64%, indicating their dependency on the debt market, as they rely on the debt market for almost 44.64% of their total assets. The mean values of MTB, RETA, and R&D are 3.590, 0.137, and 0.003, respectively.

The results of the correlation analysis are presented in Table 4. The correlation matrix provides useful information on the degree and nature of the associations among the variables under study. The absence of multicollinearity is an important assumption of multiple regression models, and according to Ref. 122, multicollinearity exists if the correlation among variables exceeds the cutoff point of 0.80 (80%). The results indicate that the correlation among variables ranges from -0.006 (between leverage and CSR) to 0.528 (between RETA and ROA), showing low-to-moderate levels of correlation. The highest value of correlation is 0.528 (52.8%) between RETA and ROA, which falls below the cutoff point; hence, there is no multicollinearity problem in our model. Moreover, the correlation coefficients of CGI and CSR with DPR were both significant and negative. The coefficient of the correlation between CSR and CG appears to be significantly positive.

To investigate the mediating role of CSR in the link between dividend policy and corporate governance, we use the mediation approach suggested by Baron and Kenny. ¹²⁰ The results of the mediation analysis are sequentially presented in Table 5.

In Table 5, Model I shows the results for the first condition of mediation, in which the independent variable CGI had a significant impact on the dependent variable DPR in the absence of the mediator variable. Model II shows the results for the second condition of mediation, revealing that the independent variable CGI significantly affects the mediator variable CSR. Model III shows the results for the third condition of mediation, in which the mediator variable significantly affects the dependent variable DPR. Finally, Model IV shows the results for the fourth condition of mediation, in which the effect of the independent variable on the dependent variable decreases and becomes insignificant after the inclusion of the mediator in the model. This is observed from the beta coefficients and p-values of (-.0300364/0.029 and -.0114956/0.520) contained in model I (DPR) and model IV (DPR), respectively.

In model I, our independent variable CG Index is found to have a significant impact on the dividend payout ratio, which confirms the first condition of mediation. The coefficient of the CG index is significantly negative (-.0300364, p<0.05), supporting H1, which states that firms with strong CG practices tend to pay low dividends. The results of Model I provide evidence in favor of the substitution hypothesis of agency theory that dividends act as a substitute for weak corporate governance. In Chinese firms with weak governance, minority shareholders are likely to face strong expropriation from majority shareholders, which results in agency conflicts between them and, hence, the associated agency cost. Dividend policies can help mitigate agency costs. ¹²³ Moreover, in firms with poor investor protection, a dividend policy is important for building a strong image by maintaining good relationships with investors. ^{25, 100} Therefore, weak CG firms tend to pay high dividends, and hence, in the weak shareholder rights scenario, high dividends act as substitutes for weak governance. ⁴² However, firms with strong governance are less exposed to these expropriations and experience low agency costs; therefore, they do not need to pay high dividends, indicating an inverse link between dividend policies and corporate governance. Our results are consistent with the findings of previous studies supporting the substitution effect of dividends that firms with high-quality governance tend to pay less dividends (see eg. Refs., 42, 45– 48, 51). Among the control variables, firm size, MTB, leverage, and R&D are significantly and negatively related to the dividend payout ratio (see e.g. Refs. 36, 112) while ROA showed an insignificant negative relationship with the dividend payout ratio, consistent with the findings of Refs. 35 and 50.

In Model II, the coefficient of the CG index is highly significant and positive (9.000815, p<0.001), indicating that firms with good CG tend to make more CSR investments. These results are in accordance with stakeholder theory, which states that apart from shareholders, the consideration of the interests of other stakeholders (e.g., environment, supplier and customers, community, and employees) is critical for corporate success. ¹²⁴ Moreover, our results are in line with the conflict resolution hypothesis, which suggests that firms with effective governance tend to be more socially responsible for resolving conflicts of interest among stakeholders because different stakeholders may have different objectives. ^{92, 125} If these conflicts remain unresolved, they may degrade firm performance. Therefore, in firms with effective corporate governance, managers are more likely to engage in CSR activities to resolve conflicts and maximize shareholder wealth. ^{126, 127}

Our results for Model II are consistent with those of Refs. 67, 93 and 95. The significant impact of the CG index on the mediator variable CSR confirms the second condition of mediation. Among the control variables, size is significantly and positively related to dividend payout ratio, while MTB, leverage, and R&D are significantly and negatively related to dividend payout ratio. ROA and RETA remain insignificant.

Model III shows that CSR has a significant negative impact on the dividend payout ratio (-.0020641, p<0.001), indicating that firms with CSR engagements tend to make fewer dividend payments. The results of model III provide evidence for hypothesis H2a and favor the equity cost of the capital channel, which suggests that firms’ CSR involvement reduces the cost of equity capital because these firms are more likely to hoard or invest cash instead of paying dividends. As socially responsible firms enjoy a good reputation, ^{63, 64} have good relationships with stakeholders and are financially less constrained, ^{58, 71} therefore they may have more incentives to hold or invest cash instead of paying dividends. An alternative explanation for the inverse relationship between CSR and dividends, based on the residual theory of dividends, might be that CSR firms tend to pay fewer dividends because CSR investments are costly, so nothing is left to pay as dividends. ^{56, 59} Findings in Model III confirmed the third condition of mediation.

Finally, Model IV shows the impact of CG on dividend policy in the presence of CSR (the mediation effect). The coefficient of the CG index is negative but insignificant (-0.0114956, p>0.05), whereas the coefficient of CSR is significantly negative (-.0020599, p<0.001). Moreover, the value of the CG index coefficient (-.0114956) in Model IV decreased from its value (-.0300364) in Model I, which means that the causal impact of CG on dividend payout ratio is significantly influenced by the presence of CSR. These findings confirm the fourth condition of mediation and provide evidence in favor of our mediation hypothesis H3. The decrease in the coefficient of the CG index in Model IV indicates that CSR mediates the link between CG and the dividend payout ratio. Our results in Model IV are in favor of our second argument about the mediating role of CSR in corporate governance- dividend policy association, which states that better CG leads to high CSR involvement, which in turn reduces dividend payments because of a firm’s preference to hoard cash or invest it instead of paying dividends; hence, CSR mediates the link between CG and dividend policy.

4.4.1 Alternative measure of dividend policy

To check the robustness of our main results on the mediating role of CSR in corporate governance–the dividend policy link–we used a dividend dummy (DD) as a proxy for dividend policy. ^{42, 118} Our dividend dummy takes the value of 1 if the firm is a dividend payer and zero if the firm does not pay dividends.

Table 6 shows the results of the logistic regression models in which the dividend dummy (DD) was used as a dependent variable. Models I, II, III, and IV provide the results for the first, second, third, and fourth conditions of mediation, respectively. In Model I, the CG index was found to have a significant negative impact (-.0927383, p<0.05) on dividend policy, confirming the first condition of mediation in the main model. The results of Model II indicate that the CG index’s impact on CSR is highly significant and positive (7.688139, p<0.001). In Model III, the coefficient value of CSR is -.0689057, with p<0.001 showing a highly significant negative impact on dividend policy. Model IV indicates that the coefficient of the CG index became insignificant after the inclusion of mediator CSR, again supporting our second argument on the mediating role of CSR in the corporate governance-dividend policy link. Overall, the results of the logistic regression analysis are consistent with our main findings presented in Table 5 and support the mediation effect of CSR on the corporate governance- dividend policy relationship.

4.4.2 Sobel test

We conducted a Sobel test of mediation to evaluate the significance of the mediation effect of CSR and confirm our main results on the mediating role of CSR that CSR carries the influence of CG on dividend policy. The Sobel test is used to test the statistical significance of a mediation effect ¹²⁸ and is helpful in determining whether there is a significant reduction in the independent variable’s effect on the dependent variable after the inclusion of the mediator. ¹²⁹ Moreover, the Sobel test is appropriate for large samples, and our sample size is quite large. Therefore, using the Sobel test to confirm the mediation effect is applicable in our study. The results of the Sobel test presented in Table 7 confirm the significance of CSR’s mediation effect on the link between CG and dividend policy. Hence, the results of the Sobel test also provide further support for our main finding that CSR influences CG on dividend policy.

4.4.3 Endogeneity bias

In accounting research, endogeneity is a common problem that can arise due to omitted variables, explanatory variables, and other instantaneous consequences. ^{130, 131} Until now, we have explored the one-way effect of CG on dividend policy. Reverse causality is implausible because any change in CG requires shareholder approval, and managers consider the governance structure as predetermined. ¹³² On the other hand, dividends are generally determined at managerial discretion. Hence, the former decision is more important than the latter. Thus, causality is likely to be from CG to dividend policy, as reported in the literature (eg. Refs. 10, 36, 123, 133). However, CG and dividend policy may be endogenously determined, and reverse causality may occur from dividend policy to corporate governance. Therefore, to address possible endogeneity bias/reverse causality, we use the GMM model to re-estimate the main model. The results of the GMM model are presented in Table 8, which indicate that there is no endogeneity bias because the p-values of the Sargan test, Hansen test, and AR 2 are all insignificant. Moreover, the signs and coefficients are similar to those of the main models.

The omitted variable bias may also cause another type of endogeneity. The link between CG quality and dividend payouts is deemed to be spurious because both can be simultaneously determined by omitted variables. ¹⁰ To control for simultaneity, we used the lagged values of the predictor variables to re-estimate the main model. The results are presented in Table 9, which are consistent with our main findings in Table 5. Hence, our results are insensitive to endogeneity.

Not applicable.