Introduction

BP’s spectacularly gaffe-prone chief exec Tony Hayward is finally getting his walking papers for his woeful mishandling of the colossal Gulf of Mexico oil spill … He is expected to be replaced by Bob Dudley, the senior U.S. executive who is handling the spill response. (Alpert 2010)

Angelo Mozilo, the CEO of mortgage lender Countrywide Financial, stepped down in July as Bank of America took over the company he co-founded in 1969. Countrywide was widely blamed for the reckless lending that later caused so many securities backed by the loans to go bad (Bloomberg Businessweek 2008).

Corporate social irresponsibility (CSiR) refers to corporate activities that negatively affect the long-term interests of a wide range of stakeholders (Strike et al. 2006), whereas corporate social responsibility (CSR) refers to firms’ perceived duty to engage in initiatives that go beyond what is required by law in response to stakeholders’ demands (e.g., McWilliams and Siegel 2000; Wood 1991). So far, existing research has focused predominantly on the implications of positive CSR practices (e.g., Waddock and Graves 1997; Orlitzky et al. 2003; Greening and Turban 2000; Choi and Wang 2009) and has paid little attention to the consequences of CSiR in organizations. Although there are many anecdotal examples of large-scale CSiR (e.g., the BP, Enron, Tyco, and WorldCom/MCI cases), the effects of such behavior across firms have not been investigated systematically. This study advances our understanding of the consequences of corporate social irresponsibility when firms fail to look after or even work against the interests of their stakeholders. The aim of this study is to understand how the occurrence and nature of strategic leadership change might be affected by CSiR.

Stakeholder theory argues that it is in the best interests of firms to invest in socially responsible actions and provide benefits to a broad range of stakeholders aside from stockholders (e.g., McGuire et al. 2003). It is widely recognized that behaving in a socially responsible manner has become a critical source of legitimacy for firms (Chiu and Sharfman 2011; Hall 1993). Management scholars and firm stakeholders, including shareholders, have started to pay greater attention to CSR because of its links to economic efficiency and market performance (Godfrey et al. 2009; Hillman and Keim 2001; Orlitzky et al. 2003). A firm that is perceived as socially irresponsible is likely to experience a loss of reputation, which may damage its long-term competitive advantage (Choi and Wang 2009) and likelihood of survival (Rao 1994). From an instrumental perspective (Aguilera et al. 2007), firms have strong incentives to rectify previous strategic errors (or ignorance) associated with CSiR. Because top management team members, particularly chief executive officers (CEOs), are responsible for the best interests of not only shareholders but also other stakeholders in their firms (Donaldson 1999), we propose that CEOs who are unable to develop and implement effective CSR strategies are likely to face a greater employment threat.

The primary purpose and central contribution of our study is an examination of whether high levels of CSiR increase the likelihood of CEO turnover. As the replacement of a CEO is a crucial event in the history of an organization (e.g., Finkelstein and Hambrick 1996), the existing literature on executive succession has focused primarily on strategic leadership change following poor corporate financial performance (e.g., Kesner and Sebora 1994; Huson et al. 2001; Ocasio 1994; Datta and Guthrie 1994). However, as CSR has become more important for firms to gain legitimacy and competitive advantage (Choi and Wang 2009; Rao 1994) and shareholders are paying increased attention to its implications for market performance, understanding the relationship between CSiR and strategic leadership change is critical. In addition to examining the likelihood of CEO turnover following CSiR, additional contributions of our study are that it predicts the nature of succession (non-voluntary versus voluntary) and successor origin (internally promoted versus externally recruited) when firms are in a CSiR state.

Further, we identify contextual boundary conditions that can strengthen or weaken the relationship between CSiR and CEO succession. As Zald (1970) stated, “The outcome of succession choices may, under specific external and internal conditions, have a large impact on organizational directions and policies” (p. 245). We focus on firm visibility to stakeholders and industry dynamism as they reflect both internal and external legitimacy pressures, as well as the risks facing firms. Firms have different levels of legitimacy pressure in society, based in part on their visibility to various stakeholders (Chiu and Sharfman 2011). Those with high visibility are likely to experience stronger stakeholder scrutiny and thus are more likely to face a “legitimacy penalty” as a result of CSiR. In addition, a dynamic industry environment is characterized as highly uncertain and unpredictable (Pfeffer and Salancik 1978), which adds to CEOs’ challenges in identifying and satisfying the interests of their stakeholders. We believe that incorporating the moderators of firm visibility and industry dynamism will help us to understand more completely the dynamic relationship between CSiR and CEO change.

The final part of our theoretical contribution is that we link CSR’s legitimacy-seeking elements to a firm’s overall risk management strategy more effectively (Husted 2005). As we argue, firms face increasing internal and external stakeholder CSR demands, and those demands are more salient to firm performance and survival. Therefore, the purpose of firms’ legitimacy-seeking behavior in the face of CSiR through strategic leadership change is to address these demands and to reduce the overall risk profile of the firm, which is at the core of the firm’s strategic mission (Bettis 1983). As Husted (2005) suggested, “CSR projects provide a way of reducing the downside risk of a firm” (p. 176). Engaging in CSiR may intensify a firm’s unsystematic risk from factors such as increased regulatory scrutiny, potential litigation, labor unrest, or environmental degradation (e.g., Orlitzky and Benjamin 2001). Each of the moderators we examine plays a role in risk management; i.e., more visible firms and those operating in more dynamic industrial environments face more risk from negative stakeholder outcomes. By linking CSiR to firm risk management, we provide an additional theoretical rationale for why firms choose succession strategies when faced with increased risk brought on by CSiR. Our study extends the implications of strategic leadership change to the social domain and enriches CSR governance research by linking the relationships among individuals (CEOs), organizations (CSR performance and firm visibility), and environmental conditions (dynamism) to better understand the consequences of firms’ CSR behavior.

Theoretical Background

CSR and Corporate Legitimacy

Before we delve into our theoretical model, it is critical that we clarify the relationship between CSR and CSiR. We conceptualize CSiR as a separate but related construct within the broader CSR space. While CSR has been studied more often, there is a growing body of literature within the broad CSR rubric that has examined CSiR as a distinct construct. Several CSR scholars have also studied CSiR (e.g., Fombrun et al. 2000; Muller and Kräussl 2011; Strike et al. 2006). For us, one of the clearer definitions of the distinction between the two constructs occurs in Tang et al. (2015):

While the positive aspect of CSR may be considered as a corporate expense that diverts valuable resources to activities that do not directly enhance shareholder value, the negative aspect of CSR is considered to be a cost-saving strategy that improves firm performance at the cost of reduced stakeholder value (p. 1342).

The key in the above definition is that CSR is a proactive effort on the firm’s part to increase stakeholder value. Alternatively, CSiR reduces stakeholder value by diverting resources away from those efforts that address stakeholder demands. This redirection is not necessarily the same as intentionally behaving badly. For example, in the case of a Bangladesh factory collapse in 2013 that killed more than 1500 people, the owner of the factory was ignorant of the garment workers’ safety issue because he put shareholders’ and buyers’ interests before labor interests. This is an act of CSiR, but likely not one where the owner intentionally put the lives of workers at risk.

According to stakeholder theory, firms should be motivated to gain legitimacy by investing in socially responsible activities for their internal stakeholders (e.g., employees) and external stakeholders (e.g., customers, suppliers, distributors, regulatory authorities, etc.) (Hillman and Keim 2001; Wang et al. 2008). Legitimacy-seeking behavior is generally motivated by legal, regulatory, or contractual actions (coercive isomorphism), value-driven dynamics (normative isomorphism), or the behavior of industry leaders (mimetic isomorphism), all of which increase a firm’s risk from stakeholders in its institutional environment if not addressed (DiMaggio and Powell 1983). As legitimacy is an essential condition for corporations to survive and prosper (e.g., Rao 1994; Singh et al. 1986), it is an important force behind firms’ CSR strategies, which form part of their legitimacy-seeking behavior in the business environment (Suchman 1995). However, we suggest that if CSiR has reached a high level, then the firm’s legitimacy and its “social license to operate” (e.g., Hartman et al. 2007) are in jeopardy. The concept regarding the “social license to operate” has become quite pervasive in the practitioner literature as well (e.g. Forbes–Maidment, 2010 or The Guardian–Morrison 2014) and suggests that not only do firms require support and approval from stakeholders to thrive and prosper but, in extreme circumstances, sufficient damage to legitimacy can hamper or eliminate the firm’s ability to operate.

Several scholars have argued that CSR investments generate positive effects by developing critical human resources and technological competencies (Russo and Fouts 1997), boosting moral capital through the fulfillment of a deontic obligation (cf. Kant 1964) that firms have to society, improving corporate image (Fombrun 1996), enhancing firm relationships with customers (Lev et al. 2010) and government (Wang and Qian 2011), and reducing litigation risk (Koh et al. 2014). These benefits of CSR investment thus provide firms with not only legitimacy and reputation (Fombrun et al. 2000) but also insurance-like protection to mitigate the potential risks from their environment (Godfrey 2005; Williams and Barrett 2000). From the resource-based view, being a good corporate citizen through CSR practices also helps a firm to achieve a sustainable competitive advantage (Choi and Wang 2009), because a strong CSR reputation cannot be imitated easily and is not substitutable by competitors (e.g., Barney 1991). Accordingly, firms that are seen as socially irresponsible are more likely to risk the loss of legitimacy and reputation from their stakeholders (e.g., Deephouse 2000; Fombrun 1996), threatening their long-term competitive advantage in the market.

CSiR and CEO Leadership Changes

Firms’ legitimacy crises due to CSiR reflect strategic leaders’ failure to satisfy firm stakeholders, including shareholders, through effective CSR strategies. As Donaldson (1999) argued, managers are at the center of stakeholder theory, suggesting the importance of corporate executives (in particular, CEOs) who decide how firms should respond to the best interests of stakeholders (Manner 2010). Increasingly, CEOs are held accountable for setting appropriate strategies that benefit the majority of firm stakeholders, not just shareholders (Hosmer 1994). More recently, Lan and Hercleous (2010) asserted that “[n]on-shareholder constituencies have stakes in the corporation that are as equally important as those of shareholders. Managers and directors should be sensitive to stakeholders’ interests when making decisions” (p. 298). Past research on top managers’ roles in firms’ CSR performance has focused mostly on executive gender (Manner 2010), compensation (Deckop et al. 2006), or background and experience (e.g., Slater and Dixon-Fowler 2009) as antecedents for CSR activities. More recent research has examined the effect of executive psychological bias, such as CEO hubris, on firm engagement in CSR (Tang et al. 2015). However, CEO succession as a consequence of CSiR has not received much scholarly attention.

Although CSR has been examined closely in terms of its relationship with a firm’s financial performance and competitive advantage (Muller and Kräussl 2011; Waddock and Graves 1997; Wang and Qian 2011), it is surprising that we have little knowledge about how a leadership change might occur following CSiR. Given that a poor reputation due to CSiR may damage a firm’s competitiveness and survival in the market plus increase its riskiness, it is paramount for firms, especially corporate boards, to ensure that proper leadership is in place to achieve both financial and social firm sustainability. To tackle this issue, we use an instrumental perspective (Aguilera et al. 2007) to examine the consequences of CSiR on CEO employment. As CSR engagement often reflects a firm’s response to internal and external pressure (Chiu and Sharfman 2011), we also examine moderators that may influence the firm’s decision to enact a CEO leadership change due to CSiR.

Hypotheses

The Effect of CSiR on CEO Succession

We predict that firms experiencing legitimacy threats resulting from CSiR will have a higher likelihood of executive turnover. When a firm engages in poor social practices, a negative spillover effect to the CEO is likely. In general, CEOs have a profound effect on CSR strategies because they determine whether and how much a firm should respond to pressure from its various stakeholders (Donaldson 1999). CEOs benefit when their firms have a reputation for practicing good corporate citizenship, as their employment prospects are enhanced as a result of being perceived as socially responsible strategic leaders. The insurance-like protection for being socially responsible can have a positive spillover effect on a CEO’s job security. In contrast, high levels of CSiR are likely to threaten strategic leaders’ employment, as stakeholders, including shareholders, are paying increased attention to CSR practices because of their impact on firms’ “bottom line” (Muller and Kräussl 2011; Wang and Qian 2011). Although CEOs of firms facing concerns over CSiR may not always be ousted or fired immediately by the board (as many such instances are in financially distressed firms), strong criticism of CSiR might prompt CEOs to leave their current post but keep another position in the firm (e.g., chairperson or vice-chairperson). In some cases, they might take a similar position at another firm; both are considered voluntary departures (Shen and Cannella 2002). As such, the consequences associated with executive turnover following CSiR may be reflected not only in executive dismissal initiated by corporate boards. Therefore, we propose that firms engaged in more CSiR generally have a higher likelihood of CEO turnover.

Hypothesis 1a

Corporate social irresponsibility increases the likelihood of CEO turnover.

Executive succession can be either voluntary or non-voluntary. While there is no commonly agreed-upon definition, non-voluntary turnover normally refers to situations when an executive terminates his or her position unexpectedly (excluding death or health problems) without a named successor, leaves office for undisclosed personal reasons, or resigns because of scandals or performance-related issues (e.g., Zhang 2008). High levels of CSiR are likely to trigger a call for a CEO change by dissatisfied stakeholders, pressuring corporate board members, who are responsible for balancing the interests of the various stakeholders (Hung 2011), to take action as part of their fiduciary duties (Hillman et al. 2001). Accordingly, when firms face substantial concerns over their CSiR, the likelihood of non-voluntary CEO turnover increases as a result.

Hypothesis 1b

Corporate social irresponsibility increases the likelihood of non-voluntary CEO turnover.

Firm Visibility as a Moderator

In addition to examining the relationship between CSiR and CEO succession, we further investigate when the legitimacy pressure to replace the executive leadership becomes stronger or weaker. Firms face different levels of legitimacy pressure (Chiu and Sharfman 2011; Matten and Moon 2008), which influences their legitimacy-saving behavior in response to CSiR. Firms that attract more attention from the public (i.e., are highly visible to their stakeholders) are more likely to make instrumental choices by “correcting” their CSiR through strategic leadership change. Highly visible firms face more legitimacy pressures and, by implication, greater risk. They therefore face greater scrutiny of their CSR practices than low-visibility firms (Campbell 2007). Firms with high visibility are also more likely to have activist stockholders, who may use shareholder initiatives to push their board to replace the CEO leadership when the firm faces a legitimacy crisis due to CSiR.

Similarly, firms with high media exposure are likely to suffer from more widespread reputation damage from socially irresponsible practices or incidents due to their public visibility. This media visibility also should increase pressure on a corporate board to take necessary actions when the current leadership fails to live up to the stakeholders’ expectations and thus puts firm legitimacy at risk. In summary, we argue that firms that face greater public scrutiny due to their high visibility to their stakeholders are more likely to initiate CEO succession in response to CSiR.

Hypothesis 2

Firm visibility to stakeholders strengthens the positive relationship between corporate social irresponsibility and CEO turnover.

Environmental Dynamism as a Moderator

Environmental dynamism is defined as the rate of unpredictable change or the level of instability in a firm’s industries (Dess and Beard 1984; Pfeffer and Salancik 1978). Highly dynamic contexts have more lethal characteristics (Anderson and Tushman 2001) because of the uncertainties firms face. As losing legitimacy increases organizational risk and uncertainty, firms operating in a more dynamic industry setting are subject to a higher probability of losing legitimacy and reputation because the interests of their stakeholders are often more ambiguous. Customer preferences and technologies also change rapidly in these contexts (Henderson et al. 2006), suggesting that CEOs need their learning to accrue quickly in such contexts. They must develop and implement effective CSR strategies for various stakeholders rapidly as their knowledge may become quickly obsolete. Thus, less stable environments pose greater challenges for CEOs and increase the chance of a “mismatch” between the CEO’s paradigm and the external environment (Henderson et al. 2006). This is in line with prior research, which finds that leadership succession becomes critical for sustaining or improving the performance of firms operating under more unstable or turbulent environments (Virany et al. 1992). Accordingly, we argue that among all the firms facing high levels of concerns over CSiR, those operating in a more dynamic environment are more likely to enact CEO change than those in a more stable environment.

Hypothesis 3

Environmental dynamism strengthens the positive relationship between corporate social irresponsibility and CEO turnover.

Successor Origin

The literature on CEO leadership change has shown that different antecedents lead to the critical choice as to whether the CEO successor is an internal or an external candidate (e.g., Dalton and Kesner 1985; Zhang and Rajagopalan 2003). In many cases, the internal/external choice may be constrained by the availability of qualified candidates in the labor market. In addition, should the CSiR be extreme enough, potential CEO applicants might shy away from the firm because it is in too risky a position. Finally, incompetent CEOs may be forced out by an insider through internal contests for control or in a “palace coup” type situation that includes having developed sufficient support among board members to assure succession (Ocasio 1994, 1999). All of these situations might lead to an insider choice simply because an outsider is not an option.

While new CEOs, regardless of their origin, are generally motivated to make changes, insiders and outsiders may have both different options and different motivations. New internal CEOs have more in-depth firm knowledge (e.g., Zhang and Rajagopalan 2004) regarding the demands and expectations of each stakeholder group. This knowledge of the firm’s stakeholders and their CSR demands is likely to make internal CEO successors a less risky choice (and even perhaps a more “responsible” one) when firms need to restore legitimacy and reputation with their stakeholders. For example, after the Deepwater Horizon oil spill in 2010, BP’s board replaced CEO Tony Hayward with internal executive Bob Dudley because his deep knowledge of the firm’s operations and potential stakeholders affected by the crisis allowed him to react quickly to the complex situation and prevent further reputation damage to BP. This suggests that firms are likely to select a new internal CEO when dealing with a legitimacy crisis resulting from CSiR.

However, some might argue that external CEOs are less constrained by a firm’s routines (Carey et al. 2000) and dominant logic (Prahalad and Bettis 1986) and thus can start from scratch (Kesner and Sebora 1994). Unlike internal CEO candidates, external CEOs generally have a limited historical connection with the predecessor CEO and little or no involvement in the new firm’s past CSR strategies. These characteristics associated with an external CEO are critical when firms must mitigate a legitimacy crisis associated with CSiR effectively. While external CEOs are often brought into rectify their predecessors’ strategic errors associated with poor financial performance (e.g., Wiersema 1995; Kesner and Dalton 1994), when firms are in a CSiR state, they might also be more capable of bringing about changes in CSR governance and restoring stakeholder confidence than internal candidates or CEOs of non-succession firms. These arguments suggest that firms should select an external CEO when dealing with substantial concerns over CSiR. Accordingly, we propose the following competing hypotheses:

Hypothesis 4a

Corporate social irresponsibility increases the likelihood of internal CEO succession.

Hypothesis 4b

Corporate social irresponsibility increases the likelihood of external CEO succession.

Methods

Sample

We identified our sample firms from the Socrates database built by Kinder, Lydenberg, Domini & Co., Inc. (KLD) (now owned by MSCI), which is based on the Russell 3000 Index (the largest 3000 publicly listed firms in the United States). We used data from between 2001 and 2008 because this minimizes the effects of the 2007–2009 recession, which are seen from 2007 onward. The KLD database is the most widely used in CSR research due to its comprehensiveness and objectivity (Choi and Wang 2009; Hillman and Keim 2001). Hart and Sharfman (2015) and Sharfman (1996) demonstrated the validity of this KLD dataset.

We removed firms from banking and other service-type industries, as firms operating in such industries usually have data missing in several categories of the KLD database. We randomly chose 300 firms from the remaining list. Among these firms, we identified 139 CEO successions using Standard and Poor’s Executive Compensation database and firm annual proxies. After removing firms with missing data, we have 124 succession firms in the final sample. Our tests compared the succession firms to a non-succession sample. We use a matched control method based on firm industries (two-digit SIC) and the number of firms per year to compile a non-succession sample of 124 firms. To avoid potential confounding effects, a non-succession firm was not to have had a CEO change within 2 years before or 2 years after the corresponding baseline year. To check whether firm characteristics were comparable between the succession and non-succession groups, we conducted several t-tests based on the total assets, total employees, slack (cash+securities), and risk (debt-to-equity ratio) 1 year prior to succession or the corresponding year in the non-succession firms. In addition, to check whether firms’ succession events were driven by poor financial performance, we also compared firms’ return on assets (ROA; both firm and industry-adjusted measures) and total stock returns between succession and non-succession samples. The results demonstrate that the succession and non-succession groups do not differ significantly from each other based on these dimensions at the 0.05 level (see Table 1).

Table 1 Comparisons between succession and non-succession samples
Full size table

Dependent Variables

We operationalized CEO succession using a binary variable, coded 1 if a firm had a CEO change during the sampling period and 0 otherwise. Our conceptual model proposes that high levels of CSiR will increase the likelihood of CEO turnover, either voluntary or non-voluntary; thus, including both types of succession is necessary for the purpose of testing Hypothesis 1.

We measured the nature of CEO succession by classifying CEO turnover as either voluntary or non-voluntary from the 124 succession cases identified above. We engaged two research assistants to investigate the reasons behind a CEO’s departure based on the news reports from major financial news outlets, including the Wall Street Journal, Bloomberg, Dow Jones News, etc. Following the approach used by Zhang (2008) in measuring CEO dismissals, we identified 102 voluntary successions and 22 non-voluntary successions based on our sample. Among the voluntary turnover cases, there are seventy-two cases of normal retirement based on the official announcement. There are sixteen cases in which the CEO left office but kept another position in the firm. There are six cases in which the CEO accepted a similar position at another firm. There are five cases in which the CEO left office due to health-related issues. There are three cases in which the CEO left office due to mergers or acquisitions.

Among the non-voluntary turnover cases, there are fourteen cases in which the CEO resigned unexpectedly without a successor named in the announcement. There are two cases in which the CEO appointment was terminated for undisclosed personal reasons. There is one case in which the CEO took early retirement because of performance-related problems. There are five cases in which the CEO was reported ousted or fired. As a robustness check, we used another popular measure developed by Shen and Cannella (2002) in which a CEO turnover is considered non-voluntary if he or she left office (1) before 64 years of age, (2) for reasons unrelated to health issues, death, mergers and acquisitions, or a new position within the same firm. All 22 CEOs classified as non-voluntary turnover met both of these criteria, indicating high consistency between the two measures. The nature of succession was coded 1 for a non-voluntary succession, 2 for a voluntary succession, and 3 for no succession.

To measure CEO successor origin, we adopted the Zhang and Rajagoporan (2010) measure and defined new internal CEOs as those with more than 2 years of firm tenure and new external CEOs as those with two or less years of firm tenure at the time of succession. Overall, there are 96 cases of new internal CEOs, 28 cases of new external CEOs, and 124 cases of no succession in the comparison sample. A new external CEO was coded 1, a new internal CEO was coded 2, and no CEO change was coded 3.

Independent Variable

Corporate Social Irresponsibility (CSiR)

To examine the relationship between CSiR and CEO turnover, we collected firms’ CSiR prior to the occurrence of CEO succession (or one year prior to the corresponding year for the non-succession sample). Because CSR strengths and weaknesses are subject to different dynamics (McGuire et al. 2003), we summed each firm’s CSR strength and concern scores separately based on the KLD database. We then used the concern scores as a proxy for firms’ CSiR while controlling for their strength scores. Specifically, we followed previous research (e.g.,Chiu and Sharfman 2011) and measured a firm’s corporate social performance based on seven “social” categories from the KLD database: community, corporate governance, diversity, employee relations, environment, human rights, and products. Each of the seven categories contained multiple items that were rated by KLD in both the CSR strength dimension (acting socially responsible) and the CSR concern dimension (acting socially irresponsible). We summed up the ratings from all seven categories to create raw scores for the CSR strength and concern dimensions. To ensure data comparability across years, we standardized the raw scores per year for each dimension because KLD implemented the binary (0/1) rating scheme in some years and the continuous rating scheme in others that were covered in our sample timeframe. We further checked for potential outliers within the CSiR ratings in our sample. The skewness for the CSiR variable (=−0.06) is close to a normal distribution (i.e., not strongly skewed to either side) and no sample falls outside 3 standard deviations in distribution, suggesting no obvious cases of extreme values in the CSiR variable. We adopted a more conservative approach by averaging 2 years of pre-succession data, as a firm’s poor social performance in 1 year may be just a temporary phenomenon.

Firm Visibility to Stakeholders

We followed Chiu and Sharfman (2011) to estimate a firm’s visibility to stakeholders, using the numbers of firm employees, common shareholders, institutional investors, news mentions about the firm from major media outlets, and shareholder resolutions as indicators. The firm’s employee and common shareholders data came from the Compustat database. Data on institutional investors came from Thomson Reuters institutional (13f) shareholdings files. The total news mentions a firm received came from the Lexis Nexis® database. Shareholder resolutions can increase firm visibility because they reflect high levels of stakeholder awareness (O’Rourke 2003). Data on shareholder resolutions came from the EthVest database of the Interfaith Center on Corporate Responsibility. The information for the five visibility indicators was based, on average, on the 2 years prior to CEO succession or the corresponding year (for non-succession firms). We then conducted a principal factor analysis using the unrotated solution. The test extracted one factor (eigenvalue = 2.9) that accounted for 68 % of the variance, with the following loadings based on the factor structure matrix: 0.85 (firm employees), 0.78 (common shareholders), 0.83 (institutional shareholders), 0.81 (firm news mentions), and 0.51 (shareholder resolutions). The scale’s Cronbach’s alpha (0.79) indicated good internal consistency for the firm visibility construct.

Industry Dynamism

We followed Dess and Beard (1984) and estimated environmental dynamism as the volatility of industry (four-digit SIC) sales over the 5 years prior to CEO succession or the corresponding baseline year (for non-succession firms). We calculated this variable as the standard error of the regression slope over the average value of industry sales in the same period. A higher score reflected that a firm operated in a more dynamic environment. We then log transformed industry dynamism to improve data normalization and reduce the effects of potential outliers.

Control Variables

We controlled for a number of factors to rule out their potential effect on our theoretical model. First, we controlled for a firm’s pre-succession CSR strength score based on the KLD ratings, because a high CSR strength score may reduce the negative effect of poor social performance on CEO change. In addition, because poor financial performance can affect a firm’s CSR activities (Waddock and Graves 1997) and is a key factor in CEO turnover (Jensen and Ruback 1983), we controlled for firms that experienced financial distress prior to succession, based on industry-adjusted return on assets (i.e., firm ROA minus industry average ROA). Specifically, a financial distress condition was coded 1 if a firm had three consecutive years of decline in ROA relative to its industry peers and 0 otherwise (i.e., industry-adjusted ROA t−3 > ROA t−2 > ROA t−1). In addition, we used financial performance lagged 1 year before succession (i.e., ROA t−1) as an alternative control in the model but it was not significant, suggesting that 1 year of financial performance may not be a good predictor for CEO replacement. To conserve degrees of freedom, we dropped the 1 year lagged performance measure and included only financial distress (3 years of performance decline) as a control in the model.

We controlled for firm size, measured as the log value of a firm’s total assets. As stakeholders may pay greater attention to larger firms, we conducted a robustness check by including firm size as one of the indicators for a firm visibility factor. The results based on the revised visibility factor remained consistent with the final model. We decided to keep firm size as a control because it also represents the level of potential resources or slack that a firm can use for socially responsible activities. We controlled for board governance, because a diligent board is more likely to take action to reevaluate CEO leadership (e.g., Westphal 1999) when its firm faces a legitimacy crisis. We measured board governance as board independence, calculated as the proportion of independent directors on the board, and female director representation on the board, calculated as the number of female directors divided by board size. While boards of directors have a fiduciary duty to replace ineffective executives, female directors have been found to pay greater attention to firms’ CSR initiatives (Hafsi and Turgut 2013; Post et al. 2011; Coffey and Wang 1998) and thus are more likely to put pressure on the board when firms face high levels of CSiR concern. We controlled for firm risk, calculated as the ratio of long-term debt to equity (Tang et al. 2015; Bromiley 1991), because it may also influence executive job security given its link to firms’ financial and social performance (Waddock and Graves 1997). In addition, institutional ownership has been shown in prior research to be a predictor for (non-voluntary) CEO turnover (Parrino et al. 2003). Thus, we controlled for institutional ownership, measured as the percentage of shares owned by institutional owners (both mutual funds and pension funds) within the firm.

We controlled for predecessor CEO age prior to succession because CEOs approaching retirement age are more likely to leave a firm. For the non-succession firms, we used CEO age prior to the corresponding matching year. We also controlled for factors contributing to CEO power because a powerful CEO can influence a board’s decision on leadership change (e.g., Dalton et al. 1998) and the firm’s CSR governance (e.g., Agle et al. 1999). These variables include CEO firm tenure, CEO equity ownership, duality, and CEO outside board memberships. CEO firm tenure was measured as the number of years that a CEO has been employed by the firm. We coded CEO duality as 1 if the CEO also served as a chairperson of the board and 0 otherwise. CEO outside board memberships is the total number of board appointments that a CEO holds outside the firm. CEO equity ownership is the percentage of firm shares owned by the CEO. We also included year dummies to control for the effect of societal attitudes or trends toward CSR. As our research focuses on succession occurrence as an outcome, we controlled for six unique years of CEO turnover from 2003 to 2008 in our sample. Please note that 2 year dummies (2003 and 2004) were dropped from the statistical estimations due to high collinearity.

We also included firm financial slack (cash and marketable securities; Kim et al. 2008) as an additional control for the statistical models predicting the nature of succession (Hypothesis 1b), because firms with more slack resources might face less pressure to take a drastic approach (i.e., dismissals) through replacing their strategic leaders because of CSiR. In addition, to avoid multicollinearity, we did not include CEO firm tenure as a control in the statistical models predicting successor origin (Hypotheses 4a and 4b), as the operationalization of our new internal and external CEOs variable was based on CEO firm tenure. Finally, we did not include year dummies in the analyses that predict the nature of CEO turnover and new CEO origin following CSiR as these two outcome variables are less likely to be influenced by external trends in the general environment.

Statistical Methodology

We recognize that firms facing high concerns over CSiR are unlikely to become socially irresponsible through a random process; rather, some firm- or governance-related attributes might play a role. Therefore, it is important to correct for potential endogeneity due to self-selection issues in our empirical models. We first split our CSiR variable based on the median value (50 % will be classified as high CSiR firms and the other half as low CSiR firms). We then conducted a two-step Heckman procedure and included the following three predictors for the first-stage dependent variable (high vs. low CSiR): presence of a CSR-related committee, the proportion of female board members, and a firm’s ROA prior to succession. The presence of a CSR committee and representation of female directors have been found to be associated with more CSR-related activities (Hafsi and Turgut 2013; Post et al. 2011; Coffey and Wang 1998; Gill 2008; Kolk and Pinkse 2010). In addition, considerable research has examined the link between corporate financial performance and social performance (e.g., Muller and Kräussl 2011; Waddock and Graves 1997; Wang and Qian 2011). To control for potential endogeneity, at least one predictor in the first-stage should not be highly correlated with the second-stage outcome variable (Larcker and Rusticus 2010). This predictor is the presence of a CSR committee. Prior research has not documented the direct influence of a CSR committee on executive turnover. We saved the inverse Mill’s ratio scores generated from the first-stage and added it as a control in the main analysis (Table 3) to correct for any self-selection bias.

We used the mean centered values of the interaction terms to reduce multicollinearity (Aiken and West 1991). In addition, we noticed a high correlation between firm assets and visibility to stakeholders (corr = 0.64). To minimize the collinearity problem, we orthogonalized these two variables to make them mathematically independent (Golub and Van Loan 1996) before adding them to the empirical tests. We used binary probit regressions to predict the occurrence of CEO succession because the dependent variable has two outcomes. We performed a robustness check using binary logit regressions, and the results are fairly consistent.

We used multinomial logistic regressions to predict the nature of succession (Hypothesis 1b) as the dependent variable has three potential outcomes (non-voluntary succession, voluntary succession, and no successions). Multinomial logistic regression is suitable when a dependent variable has more than two categories. We again used the same statistical technique to predict the successor origin (Hypotheses 4a and b), which also includes three outcomes (external succession, internal succession, and no succession). All the empirical models were tested and reported with robust standard errors.

Results

Table 2 presents the means, standard deviations, and correlations between the variables involved in the study. We also compared the CSiR ratings between succession and non-succession samples. The average standardized rating of CSiR for succession firms is 0.06 (Min/Max: −2.1/2.37; standard deviation: 0.9), and that for non-succession firms is −0.03 (Min/Max: −3.09/1.71; standard deviation: 0.6). Table 3 displays the results of the succession likelihood based on the binary probit estimates (Models 1–5), with the first model as a control model.

Table 2 Descriptive statistics and correlations
Full size table
Table 3 Regression estimates of succession occurrence
Full size table

Hypothesis 1a predicts that higher levels of CSiR increase the likelihood of CEO succession. The mean marginal effect is 0.065 and the mean z-statistics is 1.99 (>1.96, the threshold for a significance level of 0.05) based on the probit models, which indicates that an increase in the CSiR of one standard deviation leads to a 6.5 % increase in the likelihood of CEO change. The coefficient for CSiR in Model 1 of Table 3 is positive and significant (p < 0.01). Therefore, Hypothesis 1 receives support.

Hypothesis 1b predicts the nature of succession (non-voluntary succession, voluntary succession, and no succession) following CSiR. To examine this hypothesis, we used multinomial logit models with the results shown in Table 4. In Models 6a and b, we used firms with voluntary succession as a base category to compare against firms with non-voluntary succession (Model 6a) and no succession (Model 6b), respectively. As shown in Model 6a, the coefficient for CSiR is positive and significant (p < 0.05), which indicates that firms with high levels of CSiR are more likely to have non-voluntary succession than voluntary succession. Model 6b shows no significant difference in the likelihood of voluntary succession and no succession following CSiR. Model 6c shows that the likelihood of non-voluntary succession is significantly higher than no succession (p < 0.01) at high levels of CSiR. Taken together, these results suggest that the chance of a CEO leaving the position non-voluntarily is higher than the other two scenarios (voluntary succession and no succession) when the firm faces substantial concerns regarding CSiR. Thus, we conclude that Hypothesis 1b is supported.

Table 4 Multinomial logit estimates based on the nature of CEO turnover
Full size table

Hypothesis 2 predicts that firm visibility to stakeholders strengthens the positive relationship between CSiR and CEO succession. Model 3 of Table 3 shows that the coefficient for the interaction term between CSiR and firm visibility to stakeholders is not significant. The marginal effect and z-statistic distributions for this interaction term also do not support a moderating effect, as most of the z-statistic scores fall within ± 1. Thus, Hypothesis 2 is not supported, indicating that firm visibility to stakeholders does not moderate the CSiR–succession relationship.

Hypothesis 3 predicts that industry dynamism enhances the positive relationship between CSiR and CEO succession. In Model 4 of Table 3, the coefficient for the interaction term between CSiR and dynamism is positive and significant (p < 0.05). A majority of the marginal prediction scores are above zero, and the z-statistic distributions are significant at the 0.05 level. These results provide support for Hypothesis 3. Model 5 is a full model with both the main effect and two moderators included; the results are largely consistent with those from the individual models.

Hypothesis 4a predicts that firms are more likely to experience internal succession when behaving socially irresponsibly, whereas Hypothesis 4b (a competing prediction) argues that firms are more likely to have external succession under the same conditions. As noted earlier, we performed multinomial logit regressions to test these hypotheses, with the results displayed in Table 5. In Models 7a and b, we used firms with external succession as a base category to compare against firms with internal succession (Model 7a) and no succession (Model b), respectively. The coefficient of the CSiR variable is positive and significant in Model 7a (p < 0.05), but not in Model 7b, indicating that the likelihood of internal succession when the firm is in a CSiR state is significantly higher than external succession and that the difference in the likelihood of external succession and no succession is not statistically significant. Model 7c (where no succession serves as a base category) shows that the likelihood of internal succession is significantly higher than no succession (p < 0.05) at high levels of CSiR. These results provide support for Hypothesis 4a, rejecting Hypothesis 4b.

Table 5 Multinomial logit estimates based on successor origin
Full size table

Tables 3, 4, and 5 also present the fit statistics based on the Cox and Snell R-squared and Negalkerke R-squared tests, as well as the percentage of observations correctly classified in each model; the fit statistics suggests that these regression results based on the binary probit estimates and multinomial logit estimates have good model fits.

Robustness Checks

We ran a supplementary analysis using industry-adjusted CSiR measures (i.e., a firm’s CSiR rating minus the median CSiR rating of its primary industry peers); however, the results were not significant in predicting CEO turnover. It is likely that using industry-adjusted CSiR scores based on only one industry is less appropriate for our sample which consists of many firms with a well-diversified product portfolio. Thus, for firms operating in various industry segments, using industry-adjusted CSiR measure could create a biased representation of their overall CSiR ratings. Alternatively, we ran a robustness check by including the median industry CSiR rating from the firm’s primary industry as a control. The results are essentially the same as those reported in the main analysis. To conserve degrees of freedom, we dropped this variable from the regressions. In addition, we ran the analysis using three consecutive years of CSR strengths as a control because multiple years of improved CSR strengths could provide ‘insurance-like’ protection for firms facing CSiR concerns. We coded 1 if a firm had three consecutive years of improvement in CSR strength ratings (i.e., CSR strength score in t−1 > t−2 > t−3) prior to succession (or the baseline year for non-succession firms) and 0 otherwise. These additional checks generated results that are consistent with those of our main tests.

Discussion and Conclusion

We examined what happen to strategic leaders when firms engage in CSiR. We studied the antecedents of the different succession scenarios. Given the increasing pressure that firms face to be good corporate citizens (e.g., Chiu and Sharfman 2011), it is essential that theorists and practitioners understand the effect of socially irresponsible actions on a firm. Changing CEOs is a drastic response to strategic challenges and has sweeping effects on a firm (e.g., Ocasio 1994). Even after controlling for key drivers of CEO succession (e.g., firm financial performance and CEO age), our results provide evidence that CSiR is at least part of the constellation of circumstances that are more likely to lead to this major corporate response. We conclude that CSR actions are even more strategic than perhaps previously understood.

We also found that the likelihood of internal succession increases at higher levels of CSiR. From a legitimacy-saving perspective, insiders are more familiar with the requirements that various stakeholders place on the firm and present less of a risk to these constituents. As high levels of CSiR are associated with subsequent risk (e.g., Orlitzky and Benjamin 2001), changing CEOs and selecting an internal candidate also appear to be risk management tactics, suggesting that, although CSiR requires attention, firms choose to respond in a manner that results in as little functional disruption as possible. However, it is possible that it may be difficult for firms in a CSiR state to attract external candidates, so this particular finding bears further study.

Interestingly, we did not find evidence to show that firm visibility strengthens the positive relationship between CSiR and CEO succession. Perhaps CSiR may threaten legitimacy sufficiently that additional stakeholder visibility does not heighten a board’s concerns. We found that dynamic business environments exacerbated the positive relationship between CSiR and succession. As it becomes more difficult to understand stakeholder demands in a dynamic environment, using the legitimacy-seeking tools inherent in CSR becomes increasingly challenging. Firm legitimacy is already at risk in a CSiR context, so CEO succession sends both substantive and symbolic information to stakeholders that the firm plans to do things differently.

Theoretical Contributions and Practical Implications

We make four main theoretical contributions. Our first contribution is to the succession literature. We showed that CEO turnover can be a potential outcome associated with high levels of CSiR in a firm. Like poor financial performance, poor social performance can threaten strategic leaders’ employment and wealth. Our findings increase the scope of the conditions under which firms can (and should) consider replacing their top managers. By doing so, we extend the succession literature by showing a meaningful predictive effect from what has been seen traditionally as a non-market variable, i.e., CSiR.

We extend institutional theory’s legitimacy arguments in several key ways. Historically, institutional theory has focused on firm- and industry-level effects when discussing the antecedents or consequences of legitimacy-seeking behavior. By examining the role that new CEOs play in the legitimacy-saving process when firms are in a CSiR state, we extend institutional theory to the individual (i.e., upper echelon, Hambrick and Mason 1984) level. In the process of addressing legitimacy threats from CSiR, our sample firms opted for a personnel solution that helped minimize their uncertainty and risk by choosing internal candidates for their new CEOs. We thus extend legitimacy arguments to the individual level of analysis, at which both the action of replacing a CEO and where that replacement comes from may reflect firms’ intentions to save and/or rebuild legitimacy and hence maintain their “license to operate.”

Next, we added a boundary condition to the institutional perspective on legitimacy-saving behavior through our examination of a contingent moderator. We demonstrated that the need to save and secure legitimacy varied across firms and across industries (e.g., high-dynamism vs. low-dynamism contexts). Therefore, our results provide further evidence that firm and industry differences matter in the legitimacy process.

Finally, we extend CSR theory by tightening its conceptual link with risk management perspectives. Several authors have argued that CSR reduces a firm’s risk from stakeholder activism, litigation, and unwanted government intervention, among others (e.g., McGuire et al. 1988; Orlitzky and Benjamin 2001). As CEO succession is disruptive and risky for a firm (Beatty and Zajac 1987), firms with high levels of CSiR in our succession sample opted for that risky strategy probably in part to offset the greater risk presented by poor CSR. This finding helps us to conceptualize executive succession as a risk management strategy in response to CSiR or stakeholder management issues.

Limitations and Future Research

Despite its wide use, the KLD database has limitations. One criticism is the weighting scheme of the KLD data, which treats all CSR dimensions as equal (Waddock and Graves 1997). While some evidence suggests that managers perceive the dimensions differently (e.g., Ruf et al. 1993), there is no evidence that stakeholders share these perceptions. We therefore used the equal weighting approach. Although the KLD data have some limitations, we chose to use the dataset because it is more objective and comprehensive than other CSR data sources. In addition, our results may not be applicable to institutional or country environments in which stakeholders are less concerned about firm CSR. In such contexts, firms will not face the same internal and external legitimacy risks that we describe here. Firms in such contexts are thus less likely to replace their CEOs after CSiR. Similarly, the effect of CSiR on non-voluntary turnover and succession origin may not be observed.

In terms of future research, there are several fruitful next stages for this research program, but because of space constraints, we focus on three areas. First, we believe it could be valuable to examine the attributions that both CEOs and boards make in terms of CSiR and their potential implications for the firm (e.g., Fredrickson et al. 1988; Bettman and Weitz 1983). How both the CEO and the board interpret the CSiR state could have profound effects on how they make choices and how these choices affect CEO turnover. As an example, it would be interesting to see if the CEO’s influence on the board regarding CSiR is an attempt to create career insurance and the interpretations of the board’s attributions help facilitate the linkage we discuss in H1.

Second, a potentially instructive moderator is the level of regulatory scrutiny that a firm faces. Although our measure captures general visibility to stakeholders, there is one key stakeholder, the regulator, whose influence may not be estimated. Many of the topic areas under the social responsibility rubric have compliance elements inherent in them, such as a certain amount of environmental performance, product safety monitoring, or employee responsiveness required by law. Going beyond that required compliance falls into the realm of social performance. It is conceivable that firms facing high regulatory pressure respond even more strongly to CSiR by changing their CEOs quickly as both a substantive and symbolic gesture. Further, these firms may also see CSiR as an even larger risk to their legitimacy.

Finally, we argue that CEO change is a mechanism by which firms can regain or even enhance their legitimacy when they are in a CSiR state. There may be conditions under which such change does not enhance the firm’s legitimacy and may diminish it in the eyes of key stakeholders. One condition that warrants study is when stakeholders, particularly powerful stakeholders, are sharply divided on what constitutes CSiR (Agle et al. 1999). If the CEO is replaced in this situation to address one side of the social responsibility debate, powerful stakeholders on the other side may be angry. Thus, exploring the challenges of a conflicted stakeholder base as a limit on the value of CEO change in rectifying CSiR could be very fruitful.

In conclusion, our study contributes to the succession and CSR literatures by demonstrating that CSiR creates sufficient risk for a firm to trigger a CEO change. This adds further credence to CSR as an important strategic issue. We advance the legitimacy literature by demonstrating when a contingent moderator can enhance explanatory power. Our results suggest that CSR and its attendant social performance activities are increasingly part of a firm’s strategic portfolio.