Board reforms around the world: The effect on corporate social responsibility

4.1 Data source and sample selection

We first identify countries that have experienced major board reforms following Fauver et al. (2017) and Kim and Lu (2013).¹⁴ For each reform initiative, we obtain information on the year in which a reform became effective, the types/objectives of the reform involved (e.g., encouraging board independence, promoting audit committee and auditor independence, and separating the chairman and CEO positions), and the reform approach (rule-based or comply-or-explain). Appendix B provides detailed information on the implementation years, types, and approaches of the major reforms in each country. The sample period starts in 2002, the earliest year for which we can obtain CSR rating data from the Thomson Reuters ASSET4 ESG database. The ASSET4 database has a large number of international firms, and it scores firms on financial, environmental, social, and governance dimensions based on objective, publicly available information according to Thomson Reuters (e.g., stock exchange filings, annual financial and sustainability reports, non-governmental organizations' websites, and news sources).¹⁵ We focus on CSR ratings in the environmental and social dimensions and use the average rating of both dimensions as an aggregated measure of CSR.

In our sample, all 34 countries with reforms implemented since 1998 are covered by ASSET4. However, because ASSET4 has provided CSR rating data since 2002, to ensure the existence of at least a 1-year pre-reform window to examine the changes in firms' CSR, the sample period for our treatment countries starts in 2003 (three countries, France, Singapore, and the United States, implemented reforms in 2003). Our sample period ends in 2011 because all of the countries in our sample completed their reforms by 2007. Using 2011 as the end year ensures a sufficient post-reform period to conduct the analysis. We further exclude firms from financial industries (SIC codes 6000–6411).¹⁶ Finally, we exclude firm–year observations with missing data for the control variables. Our final sample consists of 20,293 firm–year observations associated with 3,514 firms from 34 countries (12,343 observations from countries with reforms implemented since 2003 and 7,950 observations from countries with reforms implemented before 2003). Among the 34 sample countries, 15 countries with major board reforms implemented during our sample period (five of which adopted rule-based reforms) are our treatment sample (see Figure 1).¹⁷

4.2 Sample distribution and descriptive statistics

Panel A of Table 1 presents the sample distribution and average CSR rating by country. The treatment group includes countries that initiated reforms during our sample period. As discussed, countries that implemented reforms in 2002 or before do not experience any reform changes during our sample period. Thus, these countries are the benchmark group.¹⁸ Both the treatment and benchmark groups exhibit large variations in the number of firm–year observations across countries, with the United States and Japan representing the largest samples in the treatment and benchmark groups, respectively. Panel B of Table 1 reports the sample distribution and average CSR rating by year. There is an increasing trend in the number of firms included in the sample, from 784 firms (3.9%) in 2002 to 3,176 firms (15.7%) in 2011. The smaller number of firms included in our sample before 2004 is mainly due to the change in the ASSET4 company universe. The increased sample size also justifies the need for a constant sample analysis when conducting empirical tests using panel data. In Table 1, Panel A, we also present the comparative statistics of firms' CSR ratings across countries for a constant sample, in which we require a firm to appear at least once in both the pre- and post-reform periods.

Our results indicate that for treatment countries, on average, firms' CSR rating improves from 0.487 to 0.600 (difference = 0.113, t-value = 13.381) (see Panel A). In comparison, for the benchmark countries, firms' CSR ratings are 0.721 on average in the post-reform period relative to an average of 0.654 in the pre-reform period (difference = 0.068, t-value = 5.837). Thus, the results of the univariate analysis of the treatment and benchmark countries provide preliminary support for a positive effect of reforms on firms' CSR.

In Panel B of Table 1, we also observe a gradual improvement in CSR rating during the sample period, which is consistent with the increasing effort that firms devote to their CSR initiatives over time. Panel C of Table 1 reports the sample distribution and average CSR rating by industry based on the 22 industry classifications of Barth et al. (1998). Firms in the chemicals industry (#4) and the manufacturing industry (#7 rubber/glass and #11 transport equipment) have the highest CSR ratings, followed by firms in the utilities industry (#16), which is consistent with the importance of CSR activities in these industries. In contrast, firms in the insurance (#20) and services (#21) industries tend to have lower CSR ratings.

Table 2 provides the descriptive statistics of the key variables used in the main analyses. To reduce the influence of outliers, all of the continuous variables are winsorized at the top and bottom one percentile. The average CSR rating is around 0.5, which is comparable to the findings of prior studies (e.g., El Ghoul et al., 2017; Ioannou & Serafeim, 2012; Lys et al., 2015).

Next, we conduct a bivariate analysis using the Pearson correlation coefficients of the major variables used in our test. Untabulated results indicate that our variable of interest, the post-reform indicator Post, is positively and significantly correlated with the three measures of CSR: CSR_Perf (0.024), ENV_Perf (0.030), and SOC_Perf (0.014). This provides preliminary evidence supporting Hypothesis 1a that firms' CSR investment improves after board reforms. The results also show that CSR rating is positively correlated with firm size, firm age, R&D expenditures, leverage, capital intensity, institutional ownership, cross-listing, analyst coverage, and Big Four auditor. These findings are generally consistent with the literature.

5.1 Baseline results

Table 3 reports the regression results for the competing hypotheses regarding the effect of reforms on CSR. Column (1) shows that the coefficient on Post is positive and statistically significant (coefficient = 0.026, p < .01) when the overall CSR rating (CSR_Perf) is the dependent variable. Columns (2) and (3) present the results when the dependent variable is measured by environmental rating (ENV_Perf) and by social rating (SOC_Perf), respectively. Again, the coefficients on Post are both positive and significant (p < .01).¹⁹ Thus, the results in Table 3 support Hypothesis 1a that a country's implementation of reforms improves firms' CSR investment on average. This finding supports the conclusion that board reforms strengthen the monitoring role of the board and empower stakeholders/investors outside of the firms to make changes in their corporate decision-making.

A company's CSR performance can be manifested in many dimensions. To conduct an in-depth analysis, we further decompose each firm's CSR performance into five categories: environment, product, workforce, society, and customer. We use the performance score in each of the five categories as the dependent variable to estimate model 1. The regression results (untabulated) indicate significantly positive estimated coefficients for four of the five categories (only the customer category is not significant).²⁰

5.2 Alternative DID specification

Our baseline results are obtained using a staggered DID research design. This design has a key advantage: it automatically takes all firms from countries without reforms at a given time as the benchmark group (i.e., including all pre-reform observations in the reform countries). However, the various benchmark groups in different years provide a less clear picture of the composition of the benchmark sample. To examine whether our findings are robust across different model specifications, we use an alternative DID design with two alternative benchmark groups: (1) the United Kingdom, the first country to implement corporate governance reforms after 1998, and (2) all countries with reforms implemented outside of the testing period, including the United Kingdom (referred to as non-reformed countries). In this specification, we create an indicator variable, Reform, which equals to 1 for the treatment firms and 0 for the benchmark firms. For the benchmark group (either the United Kingdom or all non-reformed countries), because no reform has been implemented (i.e., all years are non-reform years), we set 2004 as the pseudo reform year for the definition of Post.²¹ By using this research design, we can more directly compare the change in firms' CSR in the treatment countries in the post-reform period relative to the pre-reform period with the change in firms' CSR in the benchmark countries during the same period.

Column (1) of Table 4 uses all firms in non-reformed countries as benchmark firms. We find that the coefficient on the interaction Reform × Post is positive (coefficient = 0.032) and significant (p < .05). As shown in column (3), we find similar results when we use only firms in the United Kingdom as benchmark firms. Next, we use the propensity score matching (PSM) approach to identify more comparable benchmark firms from non-reformed countries.²² Columns (2) and (4) of Table 4 report the multivariate regression results using propensity score matched firms from non-reformed countries and from the United Kingdom, respectively, as the benchmark. The results in both columns support the prediction of a positive effect of reform on firms' CSR rating.

5.3 Other robustness checks

We perform several other tests to ensure the robustness of our findings. First, we conduct multivariate regression tests using the constant sample rather than the full sample. This approach ensures that we are comparing the same set of firms across two periods (i.e., the pre- and post-reform periods). The untabulated results show a positive and significant coefficient on Post (0.023, significant at p < .01). Second, we compare the changes in CSR rating across two constant test windows (i.e., 1 year before and 1 year after reform). Although the sample size is much smaller due to a shorter window, this approach helps mitigate the potential concerns regarding possible confounding events, which are likely to be observed during a longer window that may also drive a change in firms' CSR rating. The estimated coefficient on Post is still significant and positive (0.041, significant at p < .01). Third, we exclude all firms from the United States (i.e., the country with the highest number of observations in the treatment group), and we find results consistent with our main inference (0.016, significant at p < .05).²³ Fourth, to further alleviate the concern regarding unequal sample sizes across countries, we use a weighted least squares (WLS) model with the number of firm–year observations per country as the weighting factor. The result of this test does not change our inferences drawn from previous tests (0.085, significant at p < .01). Fifth, given the positive correlation between firm-level corporate governance and CSR documented in prior studies (Ferrell et al., 2016; Harjoto & Jo, 2011; Jamali et al., 2008; Jo & Harjoto, 2012), we include firms' corporate governance performance score (CG_Perf) obtained from ASSET4 as an additional control variable when testing model 1, and we continue to find consistent results.²⁴ Finally, Lins et al. (2017) suggest that firms may have stronger incentives to engage in CSR activities during a period of financial crisis. Therefore, we conduct a robustness test by excluding observations in the financial crisis year (2008), and we find that our result still holds.

5.4 Tests of the parallel trend assumption and the dynamic effect of board reforms

The validity of the DID method depends crucially on the parallel trend assumption. That is, in assessing the possible changes in the CSR rating during the post-reform period, we assume that the trends in the outcome variables for the treatment and control groups during the pre-reform period are similar. To examine this assumption, we add into model 1 an indicator Post_t − 1 capturing the year before the reform, Post_0 indicating the reform year, and the two interaction terms Reform × Post_t − 1 and Reform × Post_0 (e.g., Bertrand & Mullainathan, 2003). The regression results presented in Table 5 indicate that the coefficients on these interaction terms are both insignificant, providing evidence that the parallel trend assumption holds. Moreover, we use a dynamic DID research design by decomposing the post-reform period into the first year (Post_1), second year (Post_2), and third and all subsequent years (Post_3 and above) after reforms. Across all CSR measures, we find that increases in CSR rating are evident from the second year after a country's implementation of reforms and that the positive effect of reforms on CSR appears to persist for a long horizon.

5.5 Placebo tests

To further assess the validity of the DID method, we conduct placebo tests using different pseudo reform years for countries without implemented reforms during our sample period. We find no evidence of changes in CSR rating subsequent to pseudo reform years.

5.6 Year-by-year examination

In this section, we use a year-by-year DID specification for the constant sample to test the robustness of our findings. In our sample, most of the major board reforms implemented by countries are distributed over the 2003–2006 period.²⁵ As the United Kingdom implemented the earliest reform, we use it as the benchmark country in this test. We then assign a pseudo reform year to all of the firms in the benchmark country based on the actual reform year of the treatment countries. This specification allows us to test the changes in firms' CSR rating for the treatment firms relative to the changes in CSR rating for the control firms during the same pre- and post-reform years.²⁶ We repeat this test for each year for the 2003–2006 period. The results (untabulated) clearly indicate that for most of the years (three out of four) with major reforms implemented, we observe higher CSR ratings after the implementation of reforms.

5.7 Using first reform

Some countries in our sample have more than one reform (see Appendix B for the list of countries); thus, in this section, we conduct a robustness test using the earliest identified board reform (first reform) as an alternate reform year. Panel A of Table 6 shows that the coefficient on Post (which is defined based on the year of the first reform) is consistently negative across all CSR performance measures. We find similar results in Panel B where we use the alternative DID specification. Overall, our inferences do not change regardless of the definition of the reform year.

6.1 Channel test

Our main results suggest that firms' CSR performance improves after the implementation of board reforms within the country. In this section, we explore the possible channels through which reforms could have a positive impact on CSR. Recent evidence suggests that integrating CSR performance criteria in executive compensation (referred to as CSR contracting) has become a prevalent practice in corporate governance (Flammer et al., 2019; Tsang, Wang, et al., 2020). Providing compensation incentives based on social and environmental performance can prevent executives from undertaking symbolic CSR activities and direct managers' attention toward CSR initiatives that create long-term value (Maas, 2018; Matějka et al., 2009). Therefore, we examine whether the improved CSR performance observed after board reforms results from the emerging practice of integrating CSR criteria in executive compensation.

Based on the data from ASSET4, we define the variable CSRContracting as an indicator equal to 1 if a firm's executive compensation is explicitly linked to CSR/sustainability/health and safety targets, and 0 otherwise. Then we regress CSRContracting on the indicator Post and control variables in model 1. The logistic regression results of this analysis are reported in Table 7. Column (1) of Table 7 shows that the coefficient on Post is positive and significant (0.493, significant at p < .05), suggesting that indeed firms are more likely to link executive compensation to CSR criteria after board reforms. To be consistent with the main analysis, we conduct robustness checks by excluding the United States (the largest treatment country) from our sample, and the results in column (2) also show a positive coefficient on Post (0.799, significant at p < .01). Given that CSR contracting has been found to have a positive effect on CSR performance (Flammer et al., 2019), our findings in Table 7 thus suggest that the incorporation of CSR criteria in executive compensation could be a mechanism through which board reforms improve CSR performance.²⁷

6.2 Different objectives/types of board reforms

Next, we examine whether reforms with different objectives (i.e., promoting board independence, encouraging audit committee and auditor independence, and separating the chairman and CEO positions) affect CSR rating differently. The literature suggests that all of these mechanisms are associated with higher levels of board monitoring.²⁸ Because countries differ in their objectives/types of reforms, we repeat our analysis after restricting the sample to countries implementing each type of reform (see Appendix B for a list of countries with each type of reform). In Panel A of Table 8, we continue to observe a positive relationship between CSR and board reforms covering board independence as well as audit committee and auditor independence. The results are robust based on either major reform or first reform. Panel B of Table 8 reports the results of the analysis controlling for the effect of concurrent non-board governance reforms. We perform this analysis by adding a dummy variable, Post_NonBoard, which indicates periods subsequent to the reforms with additional non-board components. We find no evidence that reforms involving non-board components have incremental effects beyond those involving board characteristics. More importantly, after controlling for the effect of non-board components, we continue to find a significantly positive association between board reforms and CSR performance.

6.3 Different approaches to corporate governance reforms

We also examine whether the effect of reforms varies across comply-or-explain or rule-based reforms. Although both reform approaches are prevalent, there are conflicting views regarding which is more effective. Some argue that the one-size-fits-all rule-based approach runs the risk of overregulation, while others opine that the comply-or-explain approach might not yield the intended effect. To test the moderating effects of these reform approaches, we use an indicator variable, Rule, to identify countries with rule-based reform, and we interact this variable with Post.²⁹ Thus, the interaction term Post × Rule captures the difference (if any) in the effect of rule-based reforms relative to comply-or-explain-based reforms on CSR rating. The results (Table 9) suggest that the effect of reforms on CSR is stronger when the reform uses a rule-based approach.

6.4 The moderating effect of CSR awareness

The literature suggests that the effectiveness of corporate governance mechanisms could be conditional on country-level institutions. For instance, Dahya et al. (2008) find that board independence, a major element of corporate governance, is positively related to firm value in countries with poor investor protection, but they do not find a significant relationship in countries with strong investor protection. Regarding CSR performance, prior studies suggest that the value of CSR initiatives is greater in stakeholder-oriented countries that place greater emphasis on CSR activities (Dhaliwal et al., 2012; Simnett et al., 2009; van der Laan Smith et al., 2010). Therefore, we expect to observe a stronger positive effect of reforms on CSR rating when investors have a higher level of CSR awareness. However, in countries where the level of CSR awareness is high, it is also likely that further improvement of CSR initiatives tends to be more costly given the high level of CSR performance already observed in the pre-reform period.³⁰ This prediction suggests a weaker effect of reforms on CSR in countries with high levels of CSR awareness.

To examine whether and how country-level CSR awareness affects the relationship between reforms and CSR, we interact Post with an indicator variable, HI_Pubaware, which captures the level of public awareness of CSR issues in individual countries (Dhaliwal et al., 2012).³¹ The results reported in Panel A of Table 10 show a negative and significant coefficient on the interaction term Post × HI_Pubaware. These findings suggest a relatively weaker effect of reforms on CSR in countries with high levels of CSR awareness. This finding also raises the concern that firms with lower levels of CSR investment have gradually succumbed to global pressure to improve their CSR initiatives over time, which might explain our finding of a positive effect of reforms on CSR. Thus, in an additional test, we create a matched sample in which treatment and control firms are matched based on their CSR investment prior to the regulatory reforms to rule out the possible alternative explanation. Finally, we also directly control for firms' prior CSR performance in our regression. Our results (untabulated) are qualitatively similar in all these tests.

6.5 The moderating effect of the legal and regulatory environment

In this section, we examine whether the positive association between reforms and CSR varies with the stringency of a country's legal and regulatory environment. It is well documented in the literature that capital market consequences possibly associated with the implementation of rules or the adoption of a mandatory requirement in a country depend strongly on the country-level institutional factors related to legal and regulatory environment. For instance, Brown et al. (2014) show that the effectiveness of accounting regulation is more pronounced in countries with stronger enforcement. Christensen et al. (2013) also find that the capital market effects associated with the mandatory adoption of IFRS largely are driven by a small number of IFRS-adopting countries with enhanced country-level enforcement. Country-level institutions related to legal and regulatory environment do not only influence the effect of mandatory requirements (such as the adoption of board reforms in a country); it can also impact the effect of voluntary decisions made by the firms (such as the CSR investment decision made by managers). For example, Cao et al. (2017) find that voluntary financial reporting tends to have a stronger effect on the cost of equity capital in countries with stronger investor protection. Following the findings of these studies, we believe it is reasonable to expect that the effect of board reforms on CSR would be stronger in countries with stronger institutions related to legal and regulatory environment.³²

We add the interaction terms Post × HI_InvPro and Post × HI_RegQual in model 1 to estimate the regression. Table 10, Panel B, shows a positive and significant coefficient on Post × HI_InvPro, suggesting that the effect of reforms is stronger in countries with higher levels of investor protection. We consistently find a significant and positive coefficient on Post × HI_RegQual, which again supports the prediction that the positive effect of reforms on CSR is more pronounced in countries with a more stringent regulatory environment. In the last column, where we include both interaction terms in the same model, we continue to find positive and significant coefficients on both interaction terms. Collectively, the results support the argument that reforms cannot exert a positive influence unless accompanied by strong legal enforcement.³³

6.6 The moderating effect of ownership

Using an international setting, Dyck et al. (2019) present evidence that institutional ownership is positively associated with firms' CSR performance. They further show that the positive effect of institutional ownership on firms' CSR performance is motivated by both financial and social returns as perceived by institutional investors. Other studies suggest that firms with more concentrated insider ownership are likely to observe a greater level of conflict of interest between controlling insiders and external shareholders (Bertrand & Mullainathan, 2003; Koirala et al., 2020; Stulz, 2005). We examine whether and how institutional and insider ownership (a contrasting measure to institutional ownership) moderates the effect of reforms on firms' CSR.

Table 10, Panel C, indicates that when we interact Post with HI_Institution (an indicator variable equal to 1 if the firm's institutional ownership is greater than the median institutional ownership of all firms in the same industry, and 0 otherwise), we find a positive and significant coefficient on the interaction term Post × HI_Institution. This finding is consistent with institutional investors putting more pressure on firms to engage in more CSR activities. Next, we interact Post with HI_Insider, an indicator variable equal to 1 if the firm's CEO ownership is greater than the median CEO ownership of all firms in the same industry, and 0 otherwise. The results show a significant and negative coefficient on Post × HI_Insider, indicating a weakened effect of reforms on CSR when firms' insider ownership is high. Consistent with previous studies, our findings suggest that the presence of agency conflict reduces managers' incentives for CSR investment.

6.7 Board reform, CSR, and future financial performance

While our main findings suggest that better corporate governance encourages firms to engage in more CSR activities, they do not indicate that CSR investment is perceived favorably by investors. Managers might expropriate corporate resources for self-interest or to cover up personal misconduct (Barnea & Rubin, 2010; Masulis & Reza, 2015; Prior et al., 2008); thus, CSR investment may not always have a positive effect on firm performance (Barnett & Salomon, 2006; McWilliams & Siegel, 2000; Orlitzky et al., 2003). Because the implementation of board reforms should strengthen the board's oversight function and reduce the agency problem, managers are expected to make better CSR investment after governance reforms to ensure proper returns to shareholders. Supporting this view, agency theorists argue that corporate governance should be designed to promote CSR initiatives only when CSR activities lead to significant benefits (Jain & Jamali, 2016; McWilliams & Siegel, 2001). This view is consistent with the conflict resolution hypothesis proposed in prior studies (e.g., Jo & Harjoto, 2012, p. 57). According to this hypothesis, if corporate governance mechanisms and CSR engagement effectively resolve conflicts between various stakeholders, firm value is expected to be positively related to both the effective governance mechanism and CSR engagement. Accordingly, in this section, we further examine whether CSR investment and firms' future financial performance are more positively correlated in countries with reforms.

Our proxy for firm value is Tobin's Q (Tobin's Q), which captures the value of CSR achieved through various channels, such as increased future cash flows due to customer and supplier royalties, and lower cost of equity capital, which reflects the lower risk perceived by investors. Following previous studies (e.g., El Ghoul et al., 2017; Servaes & Tamayo, 2013; Waddock & Graves, 1997), we measure Tobin's Q as the ratio of a firm's market value to its book value of assets, in which the market value is calculated as the book value of total assets minus the book value of total equity plus the market value of equity. Table 11 reports the results of this analysis. First, we find a positive and significant coefficient on CSR_Perf, indicating that CSR investment is positively related to future financial performance. This finding is in line with the findings of previous studies (e.g., El Ghoul et al., 2017; Tsang, Hu, & Li, 2020). We also find a positive and significant coefficient on the three-way interaction term CSR_Perf × Reform × Post. This finding suggests that relative to firms domiciled in countries without reforms implemented during the same period, firms domiciled in countries with reforms are more likely to experience a stronger relationship between CSR and future financial performance after reforms.

6.8 CSR investment versus CSR reporting

Due to the increasing importance of CSR activities to firms' financial performance, many firms voluntarily publish CSR reports to signal their commitment to socially responsible behavior (Lanis & Richardson, 2012) and reduce information asymmetry between firms and investors (Dhaliwal et al., 2011; Dhaliwal et al., 2012). There is also growing evidence that investors consider CSR information in their decisions (Clarkson et al., 2013; Elliott et al., 2014; Griffin & Sun, 2013), which suggests that the transparency of CSR information matters to shareholder value. However, other studies argue that managers can use CSR reporting for purely symbolic purposes, or “green-washing” (Hemingway & Maclagan, 2004; Kim et al., 2012; Ramanna, 2013; Weaver et al., 1999). That is, firms can disclose CSR information that deviates from their actual CSR performance. Supporting this view, Petrenko et al. (2016) find evidence that firms implement CSR activities for image management without a real focus on improving financial performance.

While our evidence supports the positive influence of board reforms on firms' CSR initiatives, the studies discussed above suggest that instead of committing to more costly and substantive CSR investment, managers may simply improve their CSR reporting practices (Ramanna, 2013; Simnett et al., 2009). Improved transparency of CSR activities after reforms presumably allows better monitoring by shareholders and discourages insiders from extracting private benefits from CSR investment, which results in more substantive CSR investment. Alternatively, managers may voluntarily disclose more CSR information that deviates from their actual CSR performance for impression management (Chen et al., 2016; Clarkson et al., 2008; Muslu et al., 2019). That is, changes in CSR reporting practices after reforms are a possible alternative channel through which reforms affect firms' CSR.

In this section, we examine whether and to what extent board reforms affect firms' CSR reporting practices. Following prior studies, we use multiple CSR reporting variables to measure firms' CSR reporting practices. These indicator variables measure (1) whether a company issues a standalone CSR report in a particular year, CSRDISC; (2) whether a firm's CSR report is prepared in accordance with GRI guidelines, CSRDISC_GRI; (3) whether a firm's CSR report is assured by an independent third party for its accuracy, completeness, and reliability, CSRDISC_ASSURANCE; and (4) whether a firm includes its CSR information in its annual report, CSRDISC_INTEGRATE. We treat all of these variables as dependent variables and test the effect of reforms on each of them using a logistic model. However, in all of these tests (the results are untabulated), we find no evidence that managers change their CSR reporting practices after reforms.