Do board monitoring and audit committee quality help risky firms reduce CSR controversies?

Abstract

This study focuses on potential inhibiting and driving factors of corporate social responsibility (CSR) controversies including board monitoring intensity and audit committee quality with a particular focus on risky firms. We draw on agency, resource dependence, and slack financial resources theories to explain this association. Using an international sample between 2002–2019 and executing fixed-effects regression and Hayes’s moderation analysis methodology, we find that risky firms tend to commit more CSR controversies. Furthermore, CSR performance, firm complexity, and indebtedness exacerbate CSR controversies, whereas larger boards mitigate them. Moreover, while board monitoring intensity and audit committee quality do not prevent committing CSR controversies in absolute terms, they alleviate risky firms' CSR controversies tendency. The findings confirm agency theory and the monitoring function of the board in mitigating CSR controversies. In line with the resource dependence theory, audit committees’ independent members and members with different skills and expertise provide critical resources that help prevent CSR controversies.

1 Introduction

Corporate social responsibility (CSR) can be considered an essential strategic tool that has value-enhancing benefits which help improve the value of the firm (Malik 2015). These benefits include financial market benefits by lowering the firm’s cost of capital and cost of debt and improving stock market returns (Dhaliwal et al. 2011; Wang et al. 2011; Ye and Zhang 2011; Flammer 2013), and product market benefits by improving brand reputation and customer satisfaction which leads to increased sales and achieving the competitive advantage (Brown and Dacin 1997; Singh et al. 2008; Lev et al. 2010; Green and Peloza 2011). Despite these significant profit-related benefits in socially responsible behavior, it remains an open question of whether CSR is of value relevance to shareholders. Indeed, existing research evidence remains inconclusive regarding shareholders' gain or loss from CSR activities (see the recent literature reviews by Huang et al. (2020). Managers will be concerned about whether engaging in CSR activities (or avoiding negative outcomes) is of value to shareholders, and they need to decide the level of CSR activities that leave shareholders best off. The challenge remains how best to deploy resources and effort to engage in CSR to maximize the benefits to shareholders and stakeholders.

The literature on the dark side of CSR is still in the early stages (Godfrey et al. 2024). Corporate social controversies arise when businesses tend to do more harm and fail to meet society's expectations (Herzig and Moon 2013; Anagnostopoulou et al. 2021). This failure is likely to occur when business activities do not meet ethical standards and hurt stakeholders' social welfare (Küberling-Jost 2021). Many companies have been trying to avoid committing controversial social and environmental acts due to their damaging implications on corporate reputation, brand image, and public trust. For example, the Volkswagen (VW) and British Petroleum (BP) cases represent an utter failure of CSR.¹ Hence, it becomes important to understand the potential hindering and driving factors of CSR controversies.² In this study, we focus on managerial engagements in controversial social and environmental activities that are damaging firms' reputations and negatively impacting shareholders' wealth rather than on managerial engagements at the minimum level of CSR to align with shareholders' interests. The latter is less likely to be a result of a failure in the corporate governance system.

Previous literature has largely focused on good corporate behavior (i.e., the positive side of CSR) and examined various determinants and consequences of this behavior. Only recently, academic literature has made attention to corporate bad behavior and documented the consequences and risk implications of corporate social irresponsibility (e.g., Kölbel et al. 2017; Lenz et al. 2017; Price and Sun 2017; Godfrey et al. 2024; Jain and Zaman 2020; Dharwadkar et al. 2021). Anagnostopoulou et al. (2021) argue that CSR and CSR controversies are two separate constructs grounded on different conceptualizations. As a result, it is important to draw attention to socially irresponsible behavior to alarm stakeholders (including shareholders) and prevent future CSR controversies. Extant research has also focused on the impact of corporate governance mechanisms, including board and audit committee (AC) attributes, on corporate social responsibility (e.g., Jo and Harjoto 2012; Ben-Amar and McIlkenny 2015; Harjoto et al. 2015; Hong et al. 2016; Jain and Jamali; 2016; Al-Shaer and Zaman 2018). However, we still know little about the impact of board monitoring and ACs on CSR controversies. Although two prior studies (Godfrey et al. 2024; Jain and Zaman 2020) investigated the association between board characteristics and CSR controversies, they focused on whether board gender diversity and governance bundle (i.e., board size, board independence, CSR committee, board gender diversity, board meeting attendance) diminish CSR controversies in the US. Hence, we extend those prior studies by using AC quality (i.e., independence and expertise) and a broader board monitoring proxy (i.e., board meeting attendance rate, audit committee quality, and board committee index drawing on audit, nomination, compensation, and corporate governance committees).³ Our international sample will also provide a global perspective of whether AC quality and board monitoring reduce CSR controversies with more generalizable results and implications. Moreover, the literature on risky firms and their likelihood to be involved in CSR controversies is still in the early stages, and the question of whether board monitoring strength and effective ACs mitigate risky firms' involvement in CSR controversies is unexplored and allows for significant contribution to the literature.

Table 1 Sample distribution

Our study focuses on inhibiting and driving factors of CSR controversies and contributes to the literature in a few ways. First, while it is generally recognized that corporate governance mechanisms can affect good corporate behavior, drawing attention to bad behavior is necessary to avert future corporate social irresponsibility and alert stakeholders. In particular, there is a paucity of research on board monitoring and AC quality, and corporate social controversies with a particular focus on risky firms. For example, Jain and Zaman (2020) examine how board composition including board independence, board gender diversity, and board CSR committee influence corporate social irresponsible activities, Godfrey et al. (2024) specifically examine the impact of board gender diversity on corporate social irresponsibility, and Dharwadkar et al. (2021) focus on a specific attribute of the board (the legal background) and its association with corporate social irresponsible incidents. These studies focus on specific attributes of the board of directors and did not explore the effect of AC quality and board monitoring strength on CSR controversies; besides, they do not focus exclusively on risky firms.

Second, the existing literature on CSR controversies is still in its infancy in examining the context-specific effect of the monitoring ability of the firm in preventing controversial social incidents (Godfrey et al. 2024; Jain and Zaman 2020; Dharwadkar et al. 2021). Hence, our study fills the gap in the literature by examining the moderation effect of AC quality and board monitoring intensity on risky firms’ social irresponsibility based on the most recent and one of the largest samples of firms and data used in CSR controversies’ literature. We cover an international sample of 45,840 firm-year observations across nine sectors and 61 countries for the years between 2002 and 2019 which reinforces the generalizability of the findings. We draw on agency, resource dependence, and slack financial resources theories to explain this association.

Third, from a methodological perspective, assuming that the monitoring function is enabled by the combination of several factors rather than a single or a few board characteristics (Ararat et al. 2015), our board monitoring intensity and AC quality proxies are based on a comprehensive set of indicators, unlike prior studies which mostly use single dimensions of the board or AC structure. We measure board monitoring intensity by taking the average of three variables (adapted and modified from Ararat et al. (2015)), board meeting attendance rate, audit committee quality, and board committee index.⁴ AC quality is calculated by taking the average of AC independence and expertise (Be´dard et al. 2004; Lee and Fargher 2018). Fourth, focusing on risky firms' corporate irresponsibility and how it could be mitigated may help firms better manage the negative situation and also help regulators be extra vigilant about these firms’ association with social irresponsibility.

Using a sample of firm-year observations between 2002–2019 included in the Thomson Reuters Eikon database and performing fixed-effects regression analysis and Hayes’s moderation analysis methodology, results show that board monitoring and AC quality do not prevent committing CSR controversies contrary to our expectations. However, risky firms tend to commit more CSR controversies. Furthermore, AC quality and board monitoring intensity alleviate risky firms’ CSR controversies tendency. Our results suggest that financial difficulties push firms to commit to more controversial CSR practices which could be driven by the lack of sufficient resources to perform in a socially responsible way. Moreover, the strength of ACs and board monitoring has a context-specific effect in preventing CSR controversies; while they do not generate a direct effect on diminishing controversies in absolute terms, they have a moderating effect on risky firms’ corporate irresponsibility. Our findings support agency theory and the monitoring function of the board and its committees indicating that they are more likely to provide fundamental control over the firm’s actions (Jensen and Meckling 1976; Hill and Jones 1992) and reduce CSR controversies. Findings also support the resource dependence theory (RDT) (Hillman and Dalziel 2003) which suggests that effective ACs comprised of independent directors with different skills and expertise help firms obtain critical resources and prevent CSR controversies. Also, in line with the slack financial resources theory (Waddock and Graves 1997), risky firms might be more inclined to engage in CSR controversies under the pressure of financial constraints.

The remainder of the paper is structured as follows. The next section provides the theoretical framework and outlines the research hypotheses. Section 3 sets out the research methodology in terms of sample, data, and model. In Section 4 we report the empirical results, and in Section 5, we draw conclusions, discuss the results, and suggest implications and future research opportunities.

2 Theoretical framework and hypotheses

2.1 Board monitoring intensity and CSR controversies

Agency theory states that corporate behavior results from organizational or managerial decision-making (Walker et al. 2019). The agency theory supports the monitoring role of the board of directors and its committees suggesting that they are more likely to provide effective monitoring and oversight role over the firm's activities (Jensen and Meckling 1976; Hill and Jones 1992). On the other hand, the RDT suggests that the board and its committees are composed of directors with different knowledge, skills, and expertise which help bring rich resources to the firm (Hillman and Dalziel 2003; Jensen and Meckling 1976; Shaukat et al. 2016; Helfaya and Moussa 2017, Al-Shaer and Zaman 2018). According to RDT, the composition of the board and its committees can act as essential resources to create value by enhancing firms' reputation and engagement with various stakeholders and easing access to resources (Jizi 2017; Al-Shaer and Zaman 2018).

Companies engaging in CSR controversies can face stakeholder backlash and legitimacy threats which require an increased demand for monitoring mechanisms (Price and Sun 2017; Jain and Zaman 2020; Dharwadkar et al. 2021). Managers may not pursue the best interests of their companies and shareholders and may be driven by their own self-interest (Jain and Zaman 2020). Such behavior could go against stakeholders’ interests which creates agency problems (Jensen and Meckling 1976; Fama and Jensen 1983). The board of directors is also accountable for irresponsible corporate behaviors (Neville et al. 2019) which may impact the resource allocation decision and stakeholders' engagement. Therefore, to reduce agency problems and the conflict of interest between managers and various stakeholders, there is a need to increase the board monitoring function which is likely to reduce corporate social irresponsibility.

It is also important to establish effective board committees that bring rich resources to the firm which are likely to reduce CSR controversies. This is because board committees focus on specific tasks and perform with greater competence and pragmatism, strengthening the board monitoring function (Kolev et al. 2019). Prior research has examined the impact of board committees’ independence on firm value drawing on agency theory and suggesting that independence increases committees’ monitoring skills. For example, Nguyen and Nielsen (2010) and Aggarwal et al. (2011) show that AC independence has a positive effect on market return, and Grove et al. (2011) and Hoechle et al. (2012) show that nomination and compensation committees' independence enhances firm performance. Moreover, existing research suggests that greater expertise of committee members and frequent committee meetings are appropriate governance mechanisms that help reduce management misbehavior and improve accounting practices (e.g., Abbott et al. 2004; Krishnan et al. 2011; Cohen et al. 2014).

Frequent board meetings will generally result in greater CSR engagement (Al-Shaer and Zaman 2018). During their meetings, the board of directors can set up the CSR agenda and allocate the necessary resources for different CSR activities (Jizi 2017), and ensure engagement with different stakeholders (Rowe et al. 2014; Jensen and Meckling 1976). Prior literature suggests that poor attendance at board meetings can be seen as a sign of ineffective monitoring (Cai et al. 2010; Jain and Zaman 2020) which can lead to more involvement in CSR controversies. Low attendance at board meetings means that directors who are missing the meetings will not be engaged in direct discussion and interaction with other board members and managers about various CSR issues (Xie et al. 2003; Cai et al. 2010; Jain and Zaman 2020). Those directors will have a limited view of the firm's activities which can motivate managerial opportunism (Min and Chizema 2018; Jain and Zaman 2020), inversely impact the board monitoring function, and increase the likelihood of committing CSR controversies. Given the foregoing discussion, we propose the first hypothesis as follows:

• H1: Board monitoring intensity is negatively associated with CSR controversies.

2.2 Audit committee quality and CSR controversies

Agency theory suggests that to achieve close monitoring of management, companies need an effective AC (Collier and Zaman 2005; Turley and Zaman 2007; Al-Shaer and Zaman 2021). ACs provide substantive oversight and disclosure related to governance (Piot and Janin 2007; Beasley et al. 2009) which tend to reduce information asymmetry and agency costs and increase the transparency of their monitoring (Al-Shaer and Zaman 2021). AC oversight role is not limited to financial reporting (Yu et al. 2016). It can ensure that various corporate strategies are getting sufficient agenda time and attention by the firm (Yu et al. 2016) which leads to enhancing firms' ethical compliance and reducing socially irresponsible practices. ACs review risks associated with CSR and monitor performance through internal control and self-assessment processes (Al-Shaer and Zaman 2018). As a result, the AC risk management duty can be improved by having high-quality ACs on the board. An effective AC is likely to be more successful in allocating resources toward CSR engagement (Al-Shaer and Zaman 2018). ACs comprising independent directors with financial expertise can attract human and relational capital, which leads to better social responsibility (Helfaya and Moussa 2017; Al-Shaer and Zaman 2018). Since corporate social irresponsibility leads to high financial implications, an effective AC is essential in determining a firm’s CSR strategies and reducing the risks of irresponsible behavior.

Prior literature shows that the financial expertise of AC members makes ACs' monitoring role more effective (Abbott et al. 2004; Aier et al. 2005). ACs with financial experts have the required skills to ensure compliance with CSR policies and reduce involvement in controversial CSR activities (Yu et al. 2016). Moreover, AC independence improves monitoring quality because independent directors are not involved in the daily management of their companies (Kabir and Thai 2017), which leads to more objective and better oversight of firms' activities. Besides, independent directors are more concerned about their reputation which makes them effective monitors (Kabir and Thai 2017). They can pay attention to long-term strategies and address the needs of all stakeholders (Michelon and Parbonetti 2012; Liu et al. 2015). As a result, they are likely to be more sensitive and responsive to CSR controversies. Given the foregoing discussion, we propose the second hypothesis as follows:

• H2: AC quality is negatively associated with CSR controversies.

2.3 Risky firms and CSR controversies

This study focuses on the dark side of CSR and investigates the likelihood that financially distressed firms engage in CSR controversies. CSR controversies can be associated with environmental scandals or social misconduct that capture investors’ attention and media coverage (Aouadi and Marsat 2018). Studies that focus on the positive side of CSR show that CSR has a negative impact on financial distress risk as it improves the financing conditions of companies (e.g., Jo and Na 2012; Mishra and Modi 2013; Boubaker et al. 2020; Monti et al. 2022). A few recent studies suggest that irresponsible social behavior is related to corporate risk (Kölbel et al. 2017; Godfrey et al. 2024). We expand the latter strand of research by focusing on distressed firms’ tendency to commit controversial CSR acts.

According to slack financial resources theory (Waddock and Graves 1997), a firm with low risk faces fewer financial uncertainties which may enhance environmental and social engagements and managerial discretion to improve CSR efforts. Engagement in CSR requires the facilitation of resources to provide products and services. Thus, CSR efforts are likely to be susceptible to the availability of slack resources which can affect managers’ discretion to engage in CSR activities (Shahzad et al. 2016; Moneva et al. 2020; Dharwadkar et al. 2021). On the contrary, financially distressed firms allocate resources to short-term projects and dedicate less funds to CSR-related issues because they lack the necessary resources to invest in CSR (Orlitzky et al. 2003). Financially distressed firms can be more fragile and susceptible to CSR controversies. Research evidence shows that financially distressed firms tend to engage in earnings management practices (Campa 2015; Li et al. 2020), and disclose misleading inaccurate non-financial information (Uyar et al. 2022). A firm's CSR engagement can provide a cover against financial uncertainties (Godfrey et al. 2009; Minor and Morgan 2011; Shahab et al. 2019). However, since engagement in CSR requires managerial discretion to initiate resources or cease them (Kang et al. 2016) according to the slack financial resources theory (Waddock and Graves 1997), a risky firm might be more inclined to engage in CSR controversies under the pressure of financial constraints. Companies may pursue engaging in reputation management by replicating their peer's behavior (Godfrey et al. 2024) to be seen as socially responsible companies. As a result, such behavior, if persists without effective monitoring, is likely to increase a firm's involvement in CSR controversies. Given the foregoing discussion, we propose the third hypothesis as follows:

• H3: Risky firms are more likely to commit CSR controversies.

2.4 The moderation role of board monitoring and audit committee quality in risky firms

Financial difficulties may push firms to commit more controversial CSR incidents which could be driven by saving more costs out of CSR activities and by a lack of slack financial resources to behave in a socially responsible way. Increasing board monitoring function and the effectiveness of ACs are likely to reduce corporate social irresponsibility incidents for risky firms. CSR engagement may increase companies' costs, putting companies at a competitive disadvantage and financial difficulties (McWilliams and Siegel 1997; Islam et al. 2021). Prior literature argues that CSR represents an agency problem as managers may be inclined to use CSR practices as mechanisms to enhance their own interests (Masulis and Reza 2015; Islam et al. 2021). Board monitoring intensity and high-quality ACs perform more critical duty in risky firms and may help to reduce agency costs and monitor management activities profoundly. Such effective corporate governance mechanisms can help promote strategic initiatives including CSR initiatives (Minichilli et al. 2009; Poletti-Hughes and Martinez Garcia 2022), and involve the relevant board and committees' wealth in guiding management efficiently (Valenti and Horner 2019; Poletti-Hughes and Martinez Garcia 2022). As a result, effective corporate governance mechanisms including board monitoring functions and the existence of effective ACs are likely to moderate risky firms' engagement in CSR controversies as they impact both monitoring and the provision of resources which are likely to mitigate corporate irresponsible incidents. Given the foregoing discussion, we propose the fourth hypothesis as follows:

• H4: Risky firms with stronger board monitoring and high-quality ACs are less likely to commit CSR controversies.

3 Research methodology

Multiple data analysis approaches are performed. Detailed information about the analysis tools with the relevant justifications is provided in this section. The empirical methodology includes data preprocessing steps, univariate approaches such as sample distribution based on frequency analysis as well as descriptive statistics, and multivariate analysis approaches such as correlation analysis and panel regression analysis. The proposed research models are examined using firm-year FE (Fixed-Effects) regression analysis and moderation analysis. Finally, the robustness section incorporates various analysis approaches.

3.1 Variables

CSR controversies were measured by two variables; one by composite ESG (i.e., environmental, social, and governance) controversies score (ESGconts), and the other is governance controversies (GOVconts) (Dorfleitner et al. 2020; Refinitiv 2021; Agnese et al. 2023a). While the former is used in the baseline analysis, the latter is adopted in the robustness tests. ESGconts (based on 23 metrics) assesses a firm's exposure to environmental, social, and governance controversies and negative events appearing in global media. Normally, the default value of the controversy measure is 0, while companies with no controversies will get a score of 100 in the data source (Thomson Reuters Eikon). However, we reversed the scale by multiplying with -1 to indicate higher score shows higher CSR controversies. Thus, our aggregate CSR controversies score ranges from -100 to 0. Scores closer to -100 indicate fewer governance controversies and scores closer to 0 indicate greater controversies. The calculation of GOVconts variable is based on the average of five governance controversies indicators such as insider dealings controversies, accounting controversies, executive compensation controversies, anti-competition controversies, and bribery corruption and fraud controversies. All these five metrics' scores range from 0 to 100; the default value of each controversy measure is 0, while companies with no controversies will get a score of 100 in the data source (Thomson Reuters Eikon). To get a composite governance controversies score, we summed up these five metrics and then scaled the sum by five. However, we reversed the scale by multiplying with -1 to indicate higher score shows higher governance controversies. Thus, our aggregate governance controversies score ranges from -100 to 0. Scores closer to -100 indicate fewer governance controversies and scores closer to 0 indicate greater controversies.

We have measured the monitoring ability of the firms by two indicators: audit committee quality (AudcomQ) and board monitoring intensity (BoardMon). While AudcomQ is calculated by taking the average of audit committee independence and expertise (Be´dard et al. 2004; Lee and Fargher 2018), BoardMon is calculated by taking the average of three variables (adapted and modified from Ararat et al. (2015)): board meeting attendance rate, audit committee quality, and board committee index. The board committee index is established by the following formula: [(audit committee + nomination committee + compensation committee + corporate governance committee) / 4] * 100. While Ararat et al. (2015) use board meeting frequency in the calculation of the board monitoring index, we use board meeting attendance rate assuming that rather than the number of meetings held, directors' participation in the meetings is influential in decision-making.⁵ Also, instead of using auditor quality, we use audit committee quality based on the audit committee independence score and audit committee expertise score (Uyar et al. 2023) proposing that the auditor is an external monitor, but the audit committee is an internal monitor and an important part of board monitoring. Lastly, as Ararat et al. (2015) did, we incorporate the board committee index. Our modified board monitoring index was recently used by Gerged et al. (2023).

Firm risk is proxied by the rolling standard deviation of return on assets (StdROA) over three years (Hoang et al. 2021) and Zmijewski Financial Score (ZFS) (Zmijewski 1984; Miglani et al. 2015; Charalambakis and Garrett 2019; Habib et al. 2020; Bravo-Urquiza and Moreno-Ureba 2021) which is calculated by the equation:

$$\begin{array}{c}ZFS\;\left(Zmijewski,1984\right)=-4.336-4.513\frac{Netincome\;after\;tax}{Total\;assets}+5.679\frac{Total\;debt}{Total\;assets}\\-0.004\frac{Total\;current\;assets}{Total\;currentl\;iabilities}\end{array}$$

(1)

Furthermore, we identified several control variables that could affect the extent of committing CSR controversies in firms such as the extent of CSR commitment (ESGscore), board size (Bsize), firm size (Fsize), current ratio (CurrentR), Leverage, Capexpend, RDexpend, free float percentage (FFP), and Worldwide Governance Indicators (WGI) (Tang et al. 2015; Haque 2017; Jain and Zaman 2020; Boubakri et al. 2021; Kılıç et al. 2021; Uyar et al. 2021; Albitar et al. 2022; Hamed et al. 2022). The reason why ESGscore is incorporated as a control variable is based on the moral licensing hypothesis which argues that top leaders accrue moral credits by implementing a CSR strategy that allows them to be less vigilant toward engaging in unethical behavior (Ormiston and Wong 2013).⁶ Larger firms are more complex, and hence, they are likely to commit more CSR controversies. Poor liquidity and high leverage might impose financial constraints on firms and may push them to engage with more controversies. Research and development expenditure is considered risky, and risk might trigger greater CSR controversies. Firms with more capital expenditures are exposed to greater public scrutiny which might cause them to avoid controversial CSR acts. Firms with larger boards host greater social board capital which might prevent them to commit CSR controversies. The free float percentage of shares denotes concentrated or dispersed ownership structure, and hence shareholder power in firms may play a role in the extent of CSR controversies. Finally, regulatory quality or weakness measured by WGI might predict controversial corporate acts.

All firm-level data were retrieved from the Thomson Reuters Eikon database. The WGI data were fetched from the World Bank (2020). Further detailed information regarding the list of the variables along with descriptions is provided in Table 12.

3.2 Sample

The sample of the study covers all firm-year observations belonging to non-financial firms included in the Thomson Reuters Eikon database for the years between 2002 and 2019. We excluded financial firms from the sample due to their divergence from non-financial firms’ characteristics. The initial year for which ESG data was available is 2002 in the database which is why the study period starts with that year. The data screening process as well as the characteristics of the research sample are examined in this section. It is very crucial to investigate the raw data and pass it through the data screening processes (Hair et al. 2019). Multiple steps are performed including cleaning and preparing the data for the analysis, winsorization, outlier detection, missing value analysis, and data imputation. The raw data is retrieved from the database which was not ready for analysis. The variables of the research are organized where the variable labels on the first row and the records follow the variable names in a spreadsheet layout. Any string values were determined and converted to numerical values. Possible typos or strings within numerical values are cleaned from the data set. The data was prepared for the forthcoming analyses. Second, some of the research variables had extreme values on the tails based on the initial descriptive statistics. Therefore, StdROA, ZFS, Bsize, CurrentR, Leverage, Capexpend, and RDexpend are subject to the winsorization step. At one percent of the two tails are replaced by the extreme values with their winsorized counterpart values. Third, the significant outlier detection phase is performed. For this analysis, minimum covariance determinant-MCD methodology (Verardi and Dehon 2010) is performed. This approach can robustify the Mahalanobis distance (Verardi and Dehon 2010). A final sample size of 45,840 firm-year observations is left for further analysis after removing the significant outliers.

Fourth, the missing values are investigated. The descriptive statistics of the missing value analysis indicated that the ratio of the missing values of some of the variables is relatively small.⁷ The ratios of the missing values of the indicated variables are significantly less than 5% which can be inconsequential (Schafer 1999).

Finally, the imputation phase is employed for the missing variables. The indicated variables with the missing values were subject to the imputation step. They were imputed using the Markov chain Monte Carlo method based on linear regression as the model type for scale variables.

BoardMon and StdROA have a high ratio of missing values. Therefore, these two variables were not subject to the imputation step. Furthermore, the authors did not want to minimize the overall sample for only these two variables. Hence, the models including BoardMon and StdROA variables were run on a smaller sample while the other models were run on the full sample.

The final sample is subject to frequency analysis based on Year and Sector. The sample distribution of the final sample is presented in Table 1.⁸ Moreover, the country-level sampling distribution indicates that there are 61 countries with 5,898 unique firms and 45,840 data points (Table 13 shows the detailed sample distribution in the Appendix section).

3.3 Research models

The research models are examined using various analysis approaches. This section includes empirical analysis approaches with relevant justifications. We adopted a firm and year fixed-effects (FE) regression analysis approach to mitigate the risk of potential omitted variable bias. Employing firm-year FE allows for the control of unobservable characteristics that may persist over time (Jiraporn et al. 2016). Moreover, the firm and year FE approach is employed to address unobserved heterogeneity between firms (Wooldridge 2010). In constructing the research model, we employ regression analysis with firm and year FE utilizing the Least Squares Dummy Variable Model approach (Gujarati 2014). The functional relationship between the dependent variable and the independent variables is expressed in Eq. (2) below.

$$Y_{it}=\beta_0+\beta_1{(X)}_{it}+\beta_2{(Controls)}_{it}+Firm\;FE+Year\;FE+\varepsilon_{it}$$

(2)

The dependent variable is ESGconts denoted as the "Y_it” term in the equation. Moreover, the independent testing variables are AudcomQ, BoardMon, and ZFS which are denoted by the “X_it” term. The control variables are ESGscore, Bsize, Fsize, CurrentR, Leverage, FFP, Capexpend, RDexpend, and WGI.

The index “i” denotes the firms as the panel variable. Also, the index “t” denotes the year as the time variable. The “ε_it” term is the regular error term.

Moreover, the heteroscedasticity-consistent standard errors using the Huber Sandwich Estimator (Huber 1967) which are also called robust standard errors are reported in the regression analysis. The robust standard errors are used to control the risk of possible heteroscedasticity issues (Wooldridge 2020).

The application of the fixed-effect (FE) analysis provides various advantages: (i) reducing the possible risk of multicollinearity and estimation bias (Baltagi 2005) (ii) controlling the omitted variable bias by evaluating the changes within the panels (firms) across time (years) (Wooldridge 2010).

3.4 Multicollinearity analysis

The variance inflation factors (VIF) are used to investigate if there is any high correlation (i.e., multicollinearity) among the independent variables of the research models. The VIF values range between 1.03 and 2.67 which are significantly less than the suggested cut-off value of 10 (Hair et al. 2019). Accordingly, multicollinearity threat does not exist among the independent variables.

3.5 Moderation roles of AudcomQ and BoardMon

The baseline analysis incorporated moderation analysis to examine the proposed hypotheses with moderating effects. The moderating role of AudcomQ and BoardMon on the relationship between ZFS and ESGconts is investigated in this section. Hayes’s (2017) moderation analysis methodology using a Stata module developed by Jose (2013) is used to test the hypotheses related to moderation analysis. The functional relationship between the dependent variable and the independent variables together with the interaction variables is formulated in Eq. (3) below.

$$\begin{array}{cc}{y}_{i}=\alpha +{\beta }_{1}{x}_{i}+{\beta }_{2}{M}_{i}+{\beta }_{3}({x}_{i}*{M}_{i})+Controls+{\epsilon }_{i}& i=1,\dots ,N\end{array}$$

(3)

The dependent variable is ESGconts represented by the “y_i” term. The testing variable of interest is ZFS denoted by the "x_i” term in the formulated equation. Moreover, the moderating variables are AudcomQ and BoardMon denoted by “M_i”; the control variables ESGscore, Bsize, Fsize, CurrentR, Leverage, FFP, Capexpend, RDexpend, and WGI.

4 Findings and results

4.1 Descriptive statistics

The summary of the descriptive statistics of the research variables is presented in Table 2. According to the obtained results, the mean value of ESGconts is (-91.22 ± 22.23), GOVconts is (-53.93 ± 6.09), AudcomQ is (68.69 ± 22.99), BoardMon is (82.08 ± 10.98), StdROA is (0.02 ± 0.03), and ZFS is (-3.11 ± 1.20). The ESGconts variable is highly skewed due to the nature of the data. Most of the observations were scored -100 implying that no CSR controversies were reported for the firm while the minority of the observations are scored closer to zero with reported controversies. Prior studies also suffer from the same skewness issue of CSR controversies proxy such as Treepongkaruna et al. (2022), Agnese et al. (2023b), and Issa (2023). Having said that we addressed this skewness issue in the robustness tests by two additional analyses: (1) forming matched samples based on Propensity Score Matching and re-running the model, and (2) running the model with the natural logarithm of ESGconts. Furthermore, the average ESGscore is 40.90, and the board size is 10.01. While the mean current ratio is 2.02, leverage and free float percentage are 56% and 77% respectively.

Table 2 Descriptive statistics

4.2 Correlation coefficients

The bivariate correlation coefficients between each pair are reported in Table 3. Pearson’s correlation analysis is used to obtain the results. The results reveal that GOVconts, AudcomQ, BoardMon, and ZFS have a significant and positive linear correlation with ESGconts (p < 0.05) while StdROA does not have a significant linear correlation with ESGconts.

Table 3 Correlation analysis

4.3 Baseline analysis results

The proposed research models are examined. We perform firm and year FE regression analysis. The results are provided in Table 4 (Columns #1, 2, and 3). The results reveal that ZFS (p < 0.01) has a significant and positive relationship with ESGconts while AudcomQ and BoardMon do not have a significant association with ESGconts. Thus, the first and second hypotheses (i.e., H1 and H2) are rejected while the third hypothesis is accepted (i.e., H3). Furthermore, firms with more CSR investment, larger, and highly leveraged are more likely to commit more CSR controversies. However, firms having larger board sizes are less likely to commit CSR controversies.

Table 4 Audit committee quality, board monitoring, and firm riskiness and CSR controversies

4.4 Moderation analysis results

The moderating role of AudcomQ and BoardMon on the relationship between ZFS and ESGconts is investigated (Table 5; Columns # 1 and 2). The coefficients of the interaction variables are negative and significant. Namely, the results reveal that ZFS*AudcomQ (p < 0.05) and ZFS*BoardMon (p < 0.01) have a significant negative relationship with ESGconts. Therefore, AudcomQ and BoardMon are significant moderators on the relationship between ZFS and ESGconts. This implies that risky firms with stronger audit committees and board monitoring are less likely to commit CSR controversies. Hence, the fourth hypothesis (i.e., H4) is accepted.

Table 5 Moderating role of AudcomQ and BoardMon

The moderation analysis is also illustrated in Figs. 1 and 2 by plotting the simple slopes at -1SD (Standard Deviation), 0SD, and + 1SD of the moderating variables (AudcomQ and BoardMon). The independent testing variable (ZFS) is a continuous numerical variable, thus the ZFS values range from -1SD to + 1SD on the x-axis when plotting the simple plots. The figures clearly illustrate that the simple slopes show a negative trend when the values of ZFS increase with the interaction of AudcomQ and BoardMon.

Fig. 1

ZFS x AudcomQ

Fig. 2

ZFS x BoardMon

4.5 Robustness tests

In this section, we conduct various additional analyses to assess the robustness of the initial results by employing alternative methodologies, alternative samples, and alternative variables.

First, the baseline research models are re-run by excluding countries with less than 10 firms (Table 4, Columns #4, 5, and 6) since the small number of firms in a country may not produce reliable results. The results confirm the baseline outputs; while firm risk exacerbates CSR controversies, audit committee quality and board monitoring do not reduce it.

Second, the moderating role of AudcomQ and BoardMon is subject to an alternative sample excluding countries with less than 10 firms (Table 5, Columns #3 and 4). The results are in line with the initial moderation analysis in which the interaction variables including ZFS*AudcomQ and ZFS*BoardMon have a significant negative association with ESGconts; thus, audit committee quality and board monitoring help risky firms reduce CSR controversies.

Third, the baseline research models are re-run with an alternative dependent variable of governance controversies (GOVconts) calculated based on five proxies as described in Table 12. Thus, GOVconts is selected as the alternative dependent variable and the baseline analysis is re-run (Table 6, Columns #1, 2, and 3). Accordingly, the results show that the ZFS has a significant and positive relationship with GOVconts (p < 0.10).

Table 6 Alternative dependent variable and excluding US firms from the sample

Moreover, the baseline research models are re-investigated by excluding US firms (Table 6, Columns #4, 5, and 6). The coefficients of AudcomQ and ZFS are significantly positive concerning ESGconts (p < 0.10 and p < 0.01 respectively). This implies that after the exclusion of the US from the sample, the monitoring ability of the audit committee quality weakens as AudcomQ was insignificant in the whole sample it is weakly positive and significant after the exclusion of the US.

Fourth, the research models with moderation analysis are re-run using an alternative independent testing variable of StdROA⁹ for firm risk (Table 7, Columns # 1 and 2). Thus, the moderating role of AudcomQ and BoardMon on the relationship between StdROA and ESGconts is investigated. The results show that StdROA*AudcomQ (p < 0.05) and StdROA*BoardMon (p < 0.01) have a significant and negative relationship with ESGconts. Therefore, AudcomQ and BoardMon are significant moderators of the relationship between StdROA and ESGconts. Hence, the outputs of the baseline moderation analysis and robustness test are consistent.

Table 7 Moderation analysis with alternative test variable and excluding US

Furthermore, the baseline research models with the moderation analysis are re-run by excluding US firms (Table 7, Columns #3 and 4). The interaction variable, ZFS*BoardMon, has a significant negative relationship with ESGconts while ZFS*AudcomQ does not have a significant relationship with ESGconts. This implies that after the exclusion of the US from the sample, the monitoring ability of the audit committee quality weakens in the sample as ZFS*AudcomQ was negative and significant in the whole sample, it is insignificant now after the exclusion of the US from the sample.

Fifth, we conducted an instrumental variable (IV) analysis using Two-Stage Least Squares (2SLS) regression to mitigate endogeneity concerns (Table 8) in both the linear models and the models with moderating effects. When performing the 2SLS regression analysis for the linear models, RSE¹⁰ is used as an instrumental variable for AudcomQ and BoardMon independent testing variables since audit committee quality and board monitoring are highly likely to be affected by the RSE. Besides, RDexpenditure which denotes research and development expenditure scaled by total assets is used as an instrumental variable for ZFS independent testing variable since RDexpenditure is highly likely to trigger firm risk (Koirala et al. 2020). The results of the first stage, second stage, Durbin-Wu-Hausman test of endogeneity, and weak instrument test are reported. Accordingly, the Durbin-Wu-Hausman test of endogeneity (The null hypothesis: the regressors are exogenous) reveals that AudcomQ, BoardMon, and ZFS are endogenous regressors, thus the 2SLS approach can be utilized. Furthermore, the weak instrument test indicates that the instruments are not weak since the value of the test statistics is greater than the cut-off value of 10 (Stock et al. 2002). Lastly, we utilized a single instrument for each independent testing variable which is acceptable to use a single instrument technically (Wooldridge 2010). Therefore, the results of the overidentifying test are not available since there exist no overidentifying restrictions in the analysis of 2SLS. The results reveal that ZFS has a significant positive relationship with ESGconts while AudcomQ and BoardMon do not have a significant relationship with ESGconts confirming the baseline results (Table 8, Panel A).

Table 8 Two-Stage Least Squares (2SLS) regression analysis

In a parallel fashion, we executed the 2SLS regression analysis for the models with moderating effects. Intangibility,¹¹ CFO_Dummy,¹² ZFS_IndMean,¹³ RD_Dummy,¹⁴ along with their interactions with the moderating variables (AudcomQ and BoardMon), were employed as instrumental variables following the suggestion by Wooldridge (2010). Given that we conduct the 2SLS for the moderating variables, there are three dependent variables in the first stage, encompassing the testing variable (ZFS), moderating variables (AudcomQ or BoardMon), and the interaction variables (ZFS*AudcomQ or ZFS*BoardMon). Comprehensive results of the first stages, second stages, and post-estimation test outcomes are outlined in Table 8, Panel B. Correspondingly, the post-estimation test results are satisfied, and the second stage outcomes align with the baseline models featuring moderating effects.

In the sixth test, we utilized the Propensity Score Matching (PSM). To address the skewness in raw ESG controversies score, we created an alternative sample using the PSM with a one-to-one matching approach (Rosenbaum and Rubin 1983; Leuven and Sianesi 2003). We matched observations with ESG controversies scores different from 100 with observations having ESG controversies scores exactly equal to 100. To perform the PSM method, we formed the treatment group comprising observations with ESG controversies scores different from 100, and we established the control group with observations having ESG controversies scores of 100. PSM enables the balancing of covariate distributions between treatment and control groups, mitigating skewness issues. Propensity scores are employed to match treated and control observations, ensuring similarity in observable characteristics between the two groups which mitigates the selection bias (Rosenbaum and Rubin 1983).

After implementing the PSM and examining the descriptive statistics of ESGconts variable in the matched sample, there are 13,712 matched observations with a mean of -70.63 and a standard deviation of 32.38. The average of ESGconts changed meaningfully, moving from -91.22 to -70.63 following the application of PSM.

The linear and moderation models are re-evaluated using the PSM-matched sample. The outcomes are presented in Table 9, and they align with the initial baseline analysis results.

Table 9 PSM-based sample

To mitigate the skewness in the distribution of ESGconts, we also applied the natural logarithm transformation to the raw ESGconts data. First, we divided the raw ESG controversies score by its median. Subsequently, we applied the natural logarithm to the resulting values to obtain Ln_ESGconts_raw to address the skewness.¹⁵ This transformation compresses larger values and expands smaller values, contributing to a more normal distribution (West 2022). We re-ran the baseline research models, introducing the natural logarithm transformation of ESGconts (Ln_ESGconts_raw). The outcomes, as presented in Table 10, align with the initial analysis results.

Table 10 Natural logarithm of the dependent variable to address the skewness

Finally, the moderating role of AudcomQ and BoardMon on the relationship between ZFS and ESGconts is re-examined using the country-industry-year fixed-effect regression analysis (Table 11). The results reveal that the product terms are significantly negative, which is consistent with the initial baseline moderation analysis results.

Table 11 Country-industry-year FE regression analysis

In consequence, the findings are largely robust to alternative sub-sample excluding the US, alternative sub-sample excluding countries with less than 10 firms, using an alternative dependent variable, moderation analysis with an alternative testing variable, country-industry-year fixed-effect, and endogeneity concerns.

5 Discussion and conclusion

Although a great number of papers focused on corporate governance mechanisms and CSR performance, studies on CSR controversies (irresponsibility) are scarce (Lin-Hi and Müller 2013; Dharwadkar et al. 2021). While studying positive CSR is essential to stimulate good corporate citizenship behavior, drawing attention to bad corporate behavior is necessary to prevent future corporate social irresponsibility and to alert stakeholders. Hence, we focus on potential inhibiting and driving factors of CSR controversies such as board monitoring, and AC quality with a particular focus on risky firms.

Our findings reveal that board monitoring and AC quality do not prevent committing CSR controversies contrary to expectations. However, risky firms tend to commit more CSR controversies. Furthermore, AC quality and board monitoring intensity alleviate risky firms’ CSR controversies tendency. Although the results are robust to most further and alternative tests, we highlight a notable point that the exclusion of the US from the sample produces contrary results for AC quality; it augments CSR concerns (although weakly) and loses its significant effect in mitigating risky firms’ controversial CSR acts. Among control variables, CSR performance, firm complexity, research and development intensity, and indebtedness exacerbate CSR controversies, whereas larger boards and engaging with capital expenditures mitigate them.

Taken together with direct results and moderating effects, the ineffectiveness of board monitoring and AC quality in reducing CSR controversies may not necessarily mean the weakness of the internal governance mechanisms but is more related to the firm-specific conditions, institutional environment, and focus of those two monitoring mechanisms. Board monitoring and AC may not prioritize CSR controversies but financial reporting quality. For example, Krishnan et al. (2011), Cohen et al. (2014), and Kusnadi et al. (2016) found that audit committee expertise and quality are drivers of financial reporting quality. Recent evidence also indicates that audit committee expertise and independence reinforce the value relevance of CSR reporting in the stock market (Kuzey et al. 2023). Moreover, board monitoring is influential in preventing corporate financial manipulation and fraud (Sharma 2004; Wahid 2019; Files and Liu 2022). However, the unexpected insignificant finding might be related to the institutional environment as well. Although our global sample did not produce a direct effect of board monitoring and AC mechanisms on reducing CSR controversies, two US-based studies (Godfrey et al. 2024; Jain and Zaman 2020) found that board gender diversity and governance bundle (i.e., board size, board independence, CSR committee, board gender diversity, board meeting attendance) diminish corporate social irresponsibility. However, we want to note that these two cited studies are conducted in the US and their governance metrics are different from the ones we used. The divergence between our results and previous studies' findings in direct effects might be associated with differing board monitoring metrics and the institutional environments in which the research is conducted. Lastly, our insignificant direct effects might signal that board monitoring and AC mechanisms may be prioritizing financial reporting quality and preventing financial fraud rather than CSR controversies.

However, the two monitors (i.e., AC and board monitoring) are extra-vigilant in harnessing CSR controversies in financially distressed firms which are more likely to commit CSR irresponsibility and hence may weaken the already fragile situation of the risky firms. Indeed, prior research points out the contingency's role in the negative effect of CSR controversies. For example, Kölbel et al. (2017) proved that CSR controversies if covered by the media evoke firm risk, and Nardella et al. (2020) found that CSR controversies alone do not trigger reputation loss unless the firm is found culpable by law. In line with these findings, shareholders' reactions to these CSR controversies are also mixed since they may react indifferently (Fisher-Vanden and Thorburn 2011; Lenz et al. 2017), or negatively (Price and Sun 2017; Godfrey et al. 2024), or positively (Aouadi and Marsat 2018; Melinda and Wardhani 2020). Consequently, this evidence might have a relaxing effect on the firm’s monitoring function about controversies' negative effect on stakeholders unless the firm is risky. Concerning the CSR controversy and firm risk relationship, there is not much evidence; in the opposite direction of the causality we investigated, Oikonomou et al. (2012) found that controversial CSR augments firm risk; hence, we find evidence for the reverse causality. Taken together, this evidence shows the existence of a bi-directional relationship between controversial CSR and firm risk.

Although positive CSR was among the control variables, we find it noteworthy to briefly comment on the positive association between CSR and CSR controversies as it is not much highlighted in the past literature. Ormiston and Wong (2013) explain this connection with moral licensing in such a way that engagement with CSR helps firm managers gather moral credits in society which eventually allows them to commit controversies that are supported by their empirical evidence.

6 Implications and future research avenues

The results suggest several theoretical and practical implications. The findings confirm agency theory and the monitoring function of the board indicating that it is highly likely to provide fundamental control over the firm’s actions (Jensen and Meckling 1976; Hill and Jones 1992) associated with CSR controversies. In line with RDT, (Hillman and Dalziel 2003) ACs reinforced with independent members and members with different skills and expertise provide critical resources in preventing CSR controversies. Besides, the results confirm the slack financial resources theory (Waddock and Graves 1997) that risky firms have a higher tendency to commit controversial CSR acts under the pressure of financial constraints.

The findings suggest several practical implications as well. It appears that the strength of AC and board monitoring strength have a context-specific effect in preventing CSR controversies; while they do not generate a direct effect on diminishing controversies in absolute terms, they have a moderating effect on risky firms' corporate irresponsibility. This could be attributable to two factors. First, ACs and board monitoring may not take CSR controversies seriously unless the firm is financially distressed. Second, considering the findings concerning the control variables, CSR concerns are natural outcomes of CSR engagement and firm complexity. Hence, intensive CSR engagement and the complexity of firms' operations induce committing CSR concerns of which ACs and board monitoring cannot further reduce those concerns or do not observe a serious drawback in these controversies.

Moreover, financial difficulties push firms to commit to more controversial CSR practices which could be driven by saving more costs out of CSR practices and by lack of sufficient funds to behave in a socially and environmentally responsible way. This finding may imply closer scrutiny of risky firms’ CSR practices by regulatory bodies such as stock market regulators. However, the implication of the moderation analysis is that internal governance and monitoring strength provide a resolution for and barrier to risky firms’ controversial activities. Despite the lack of direct effect of board monitoring and AC quality on controversial CSR, the existence of a moderating effect implies that these internal governance attributes pay particular attention to mitigating CSR irresponsibility in risky firms due to the reality that they tend to commit more CSR misconduct. However, we underline that non-US countries and firms need to review and revise their regulations and qualifications concerning ACs so that they can be more influential in reducing financially distressed firms' CSR controversies.

Although the study provides valuable insights concerning the dynamics of CSR concerns, future studies could deepen the investigation by trying several other contingencies. For example, other than board monitoring, the board structure's potential effect on diminishing controversial activities could be tested. Moreover, external governance mechanisms could play a role in dissuading firms from undertaking CSR controversies which is worth investigating to suggest policymaking implications. It is also possible to imitate the approach of the study in particular contexts such as European Union member states, emerging markets, and transition countries which are quite different contexts and may bear different insights. This further investigation is particularly justified by slightly different results between the US and the non-US samples in this study.

Appendix

Table 12 Definitions of the variables

Table 13 Country-level sample distribution