Although managers and researchers have invested considerable effort into understanding corporate social responsibility (CSR), less is known about corporate social irresponsibility (CSiR). Drawing on strategic leadership and moral licensing research, we address this gap by considering the relationship between CSR and CSiR. We predict that prior CSR is positively associated with subsequent CSiR because the moral credits achieved through CSR enable leaders to engage in less ethical stakeholder treatment. Further, we hypothesize that leaders’ moral identity symbolization, or the degree to which being moral is expressed outwardly to the public through actions and behavior, will moderate the CSR–CSiR relationship, such that the relationship will be stronger when CEOs are high on moral identity symbolization rather than low on moral identity symbolization. Through an archival study of 49 Fortune 500 firms, we find support for our hypotheses.

In 2008, the CEO of British Petroleum (BP), Tony Hayward, announced that BP's safety record, which was one of his key foci as head of BP, was “now among the best in our industry” (BP, 2008). Under his leadership, BP ran operational safety training sessions for its employees and encouraged a culture of safety in an effort to attend to key stakeholders such as employees, the community, and the environment (Kolmar, 2007). However, in 2010, the Deepwater Horizon explosion shattered BP's positive safety record when managers missed key safety warning signs, resulting in the loss of 11 lives and creating the worst offshore oil spill in United States history (Jervis & Levin, 2010; Rogers, 2010). Moreover, Hayward's reaction to the spill was seen as cold and perceived as minimizing BP's responsibility for the spill as well as the environmental damage to the Gulf of Mexico (Korosec, 2010; Zarroli, 2010). Following from such instances of corporate wrongdoing, business practitioners and academics have dedicated increased attention to firms’ corporate social responsibility (CSR), which refers to firm behavior that goes beyond compliance or legal requirements to provide some social good (McWilliams & Siegel, 2001). One prominent approach to understanding CSR is the stakeholder approach in which a firm is deemed socially responsible when its actions attend to the interests of not only its shareholders but also other firm stakeholders (Branco & Rodrigues, 2006; Carroll, 1991, 1999; Godfrey & Hatch, 2007). Drawing on this stakeholder approach, researchers have focused on understanding the antecedents and consequences of CSR. For instance, top management team (TMT) demographics, values, sociocognitive traits, and firm attributes such as decentralization of decision making and research and development expenditures have all been identified as important antecedents to a firm's ability to manage the needs of multiple stakeholders (e.g., McWilliams & Siegel, 2001; Thomas & Simerly, 1994, 1995; Waldman et al., 2006; Wong, Ormiston, & Tetlock, 2011). In turn, research indicates that, on average, CSR is positively related to firm financial performance (for a review, see Margolis & Walsh, 2003; Orlitzky, Schmidt, & Rynes, 2003). Although this research has advanced our understanding of CSR, several important questions remain. First, although we have a better understanding of CSR and its benefits (e.g., increased firm financial performance), past research has rarely considered the negative consequences of engaging in CSR. Relatedly, we have less of an understanding of corporate social irresponsibility (CSiR; Lange & Washburn, 2012), which refers to actions that negatively impact the interests of stakeholders (Strike, Gao, & Bansal, 2006). The limited research on CSiR has considered how CSiR and CSR are related, with research finding that they can occur concurrently (e.g., Strike et al., 2006) or that CSiR is offset by subsequent CSR (e.g., Chatterji & Toffel, 2010; Kotchen & Moon, 2011; Muller & Kräussl, 2011). For instance, Chatterji and Toffel (2010) found that firms that received negative environmental ratings subsequently improved their environmental performance. Although such research helps us to better understand the relationship between CSiR and subsequent CSR, it does not consider whether CSR influences later CSiR. Returning to the BP example at the beginning of this paper, this literature provides little guidance to understanding the conditions under which BP's socially responsible behavior (i.e., increasing its safety record) might have subsequently increased its less responsible behavior (i.e., ignoring safety warnings and minimizing BP's responsibility). Second, studies examining CSR and CSiR have generally provided a firm‐level perspective of this relationship. However, given that strategic leaders (i.e., CEO and the senior executives who belong to the TMT) make strategic decisions about CSR, it is surprising how few studies explore the effect of top leaders on CSR (but see Thomas & Simerly, 1995; Wong et al., 2011), and only one study to our knowledge theorizes that strategic leaders affect CSiR actions (Pearce & Manz, 2011). Further, few researchers approach CSR from a social psychological perspective (Rupp, Williams, & Aquilera, 2011), even though applying these theories could elucidate when leaders’ personal characteristics influence the CSR–CSiR relationship. Therefore, important questions remain as to whether and when leaders influence the relationship between CSR and CSiR. In this paper, we address these questions by examining the relationship between CSR and CSiR. First, we study whether prior CSR affects subsequent CSiR. We approach this question through the lens of strategic leadership, which suggests that leaders influence organizational strategy and in turn firm outcomes (e.g., Finkelstein & Hambrick, 1996) such as CSR and CSiR. Specifically, we focus on the effects of strategic leaders’ psychological processes on CSiR. We draw on moral licensing research to argue that top leaders who have previously implemented CSR strategies will glean moral credits from this past behavior, which leads them to develop a strategy that mistreats firm's stakeholders. Second, we consider the moderating role of leader characteristics on this relationship. Although moral licensing research consistently shows a positive relationship between moral behavior and subsequent immoral behavior, recent research suggests that this relationship may be affected by individual differences (Mulder & Aquino, in press; Zhong, Ku, Lount, & Murnighan, 2010). One relevant individual difference is moral identity symbolization, or the degree to which being moral is expressed outwardly to the public through actions and behavior (Aquino & Reed, 2002). We argue that leaders’ moral identity symbolization moderates the relationship between prior CSR and CSiR. Our study furthers research on strategic leadership, CSR, moral licensing, and moral identity by examining the effect of strategic leaders’ psychological processes and personal characteristics on CSiR in publicly traded firms.

The Influence of Prior Corporate Social Responsibility on Corporate Social Irresponsibility Over the past few decades, firm stakeholders, including shareholders, consumers, and employees, have put increasing pressure on firms to manage their interests; in other words, firms are increasingly expected to heed the calls of their multiple stakeholders. When firms are able to successfully manage these myriad demands, they are deemed socially responsible (Freeman, 1984). As business practitioners’ demands and academics’ understanding of CSR have become finer grained, so too have the definitions of CSR. Greenwood (2007) suggests CSR exists along a continuum anchored at one end by firms that actively engage with and act in the interest of varied stakeholders (i.e., are responsible) and anchored at the other end by firms that have more limited interactions with and may act in the interest of only powerful stakeholders (i.e., are irresponsible). Others have argued that although responsibility and irresponsibility are related constructs, they are distinct and thus should not be considered as opposite ends of the same continuum (Kotchen & Moon, 2011; Mattingly & Berman, 2006; Oikonomou, Brooks, & Pavelin, 2012; Strike et al., 2006). This more recent distinction between CSR and CSiR highlights that they both involve action toward stakeholders and that firms can potentially engage in both CSR and CSiR. We adopt this view and consider stakeholder management to involve both CSR, or actions that attend to the interests of multiple stakeholders beyond just shareholders (Branco & Rodrigues, 2006; Carroll, 1991, 1999; Godfrey & Hatch, 2007), as well as CSiR, or actions that negatively affect stakeholders (Strike et al., 2006). Examples of CSR include charitable giving within a community, providing employees flexible work arrangements such as flextime, and ensuring the quality of consumer goods. Examples of CSiR include actions that have resulted in controversies concerning the firm's economic impact on the community, gender discrimination, and designing and selling unsafe products and services (Kaptein, 2008). In short, the more positive actions toward its stakeholders, the more socially responsible a firm is deemed, whereas the more negative actions toward its stakeholders, the more socially irresponsible a firm is deemed (Strike et al., 2006). Given these distinctions between CSR and CSiR, a growing body of research has sought to understand and clarify their relationship. In a first examination of the CSR–CSiR relationship, Strike and colleagues (2006) found that international diversification led to both CSR and CSiR, suggesting that firms could simultaneously be responsible and irresponsible toward their stakeholders. Additional evidence for this relationship can be drawn from Mattingly and Berman (2006), who found that environmental CSR was positively correlated with concurrent environmental CSiR. They noted, “perhaps firms that tend to adopt environmentally friendly programs are often those that tend also to cause harm or to extract from it” (p. 34) and suggested that further research into this positive correlation between CSR and CSiR was needed. Other research finds that socially irresponsible behavior (i.e., CSiR) sets the stage for subsequent socially responsible behavior (i.e., CSR; e.g., Chatterji & Toffel, 2010; Kotchen & Moon, 2011; Muller & Kräussl, 2011). For example, Muller and Kräussl (2011) found that companies with poor social reputations were more likely to make disaster relief donations after Hurricane Katrina than were companies with positive social reputations. The key argument underlying these studies is that firms offset their negative reputation by engaging in CSR. Although this research has been instrumental in augmenting our understanding of why CSiR influences CSR, researchers have yet to explore whether the opposite relationship occurs such that CSR influences CSiR. In addition, although previous research on the link between CSR and CSiR has primarily taken a firm‐level approach, we propose that a more micro‐level approach would be useful to disentangle the linkages between CSR and CSiR. In this paper, we draw upon strategic leadership theory and psychological licensing research to examine the role of strategic leaders’ psychological processes and personal characteristics on CSiR. Below, we briefly review these approaches. With regard to the strategic leadership approach, one fundamental tenet is that leaders (e.g., the CEO and senior executives who comprise the TMT) typically make decisions regarding the firm's strategic direction, including social performance policies (Finkelstein & Hambrick, 1996; Hambrick & Mason, 1984; Thomas & Simerly, 1994; Wood, 1991). Thus, we approach our understanding of how CSR relates to CSiR through an examination of the ways in which leaders’ strategic decisions regarding CSR relate to their subsequent CSiR decisions. Research on psychological licensing can help us understand how leaders’ strategic decisions regarding CSR can influence later CSiR. Psychological licensing refers to perceiving that one can behave in a potentially socially undesirable way without fear of discrediting his or her self‐image (Miller & Effron, 2010). Whether an individual feels licensed or not is based on one's past behavior (Monin & Miller, 2001). For instance, having been a healthy eater for 11 months out of the year, one may feel licensed to indulge during the holiday season. In this way, one can still feel as though they are a healthy person because the 11 months of healthy eating bolsters one's self‐image and, therefore, diminishes the risk of discrediting him‐ or herself as an unhealthy person. The concept of licensing in moral domains (termed moral licensing) is germane to our focus on CSiR. Moral licensing is the psychological process that leads people to engage in morally questionable behavior after having previously engaged in socially desirable behavior. According to moral licensing research, people desire to maintain a positive moral image not only to others but also to themselves (Mazar, Amir, & Ariely, 2008; Monin & Jordan, 2009; Monin & Miller, 2001). Therefore, they want to avoid engaging in actions that are morally discrediting or that signal to themselves or others that they are unethical (Monin & Miller, 2001). As in the case of general psychological licensing, when determining if an action will be morally discrediting, people look to their past behavior, and if “people can call to mind previous instances of their own socially desirable or morally laudable behaviors, they will feel more comfortable taking actions that could be seen as socially undesirable or morally questionable” (Miller & Effron, 2010, p. 118). A number of studies have shown that moral behavior leads to morally questionable behavior at the individual level. In one of the first demonstrations of this effect, Monin and Miller (2001) found that people who were given the opportunity to disagree with sexist statements (e.g., “Most women are better off at home taking care of the children”) were subsequently more likely to make a sexist decision about who to recruit for a manufacturing job. Monin and Miller termed this process of non‐prejudiced attitudes (i.e., socially desirable behavior) subsequently leading to prejudiced attitudes (i.e., morally questionable behavior) as “moral licensing.” This study highlights that moral licensing is domain specific such that engaging in moral behavior in one domain leads to immoral behavior in the same domain. Although a large body of research has demonstrated the domain‐specific nature of moral licensing, recent research also shows non‐domain specific moral licensing can occur such that being ethical in one domain licenses a person to be less ethical in another domain. Mazar and Zhong (2010) show that prior moral behavior (i.e., environmentalism) can lead to morally questionable behavior (i.e., giving less to charity) as well as clearly immoral behavior (i.e., lying in order to receive a greater payout). As applied to our research, these studies are important in that, first, moral licensing is argued to have occurred when moral behavior leads to either morally questionable behavior or clearly immoral behavior, and second, it can be either domain or nondomain specific. Not only does one's own past moral behavior license someone to subsequently engage in immoral behavior, but moral licensing research finds that people use the actions of their ingroup, or similar others, as a guide for their own future behavior. Drawing on research on identity and vicarious self‐concept theory (Goldstein & Cialdini, 2007), Kouchaki (2011) suggested that people incorporate aspects of important others into their own self‐concept and use others’ moral behavior to define who they are. Consequently, when ingroup members behave in a moral manner, individuals internalize this moral behavior as their own and feel licensed to engage in morally questionable behavior. The previously discussed findings on moral licensing give rise to the question of why people feel licensed to engage in immoral behavior after previously engaging in socially desirable behavior. One reason may be that moral behavior is like a monetary currency (or creates moral credits) that can be banked and then drawn on when one engages in less moral behavior (Sachdeva, Iliev, & Medin, 2009). Nisan's (1990, 1991) moral balance model first discussed this notion of moral credits, arguing that individuals aim to maintain moral balance (i.e., balance between good and bad deeds) and use past behavior to determine their balance. After a person feels he or she has engaged in enough moral behavior to achieve moral balance, they are subsequently more likely to engage in immoral behavior. This inconsistent behavior (i.e., immoral behavior after moral behavior) is thought to stem from people's tendency toward self‐interest (desire to be less moral) on the one hand and their tendency to behave according to social norms (desire to appear moral) on the other hand (Beruchashvili, Gentry, & Price, 2006; Nisan, 1990; Sachdeva et al., 2009). In a series of survey studies, Nisan (1991) found support for the moral balance model. We draw on Nisan's moral balance argument to suggest that, after making strategic CSR decisions, leaders are more likely to engage in unethical behavior because of the moral credits they believe they have accumulated. In summary, this past research has been instrumental in augmenting our understanding of moral licensing at the individual level within experimental settings; however, researchers have not explored whether and when moral licensing occurs within organizational contexts. The importance of studying moral licensing in an organizational context can be drawn from a recent study by Castilla and Benard (2010), which examined gender discrimination. In a set of scenario studies they found that meritocratic organizational values led participants to make gender‐biased decisions about bonuses. Castilla and Benard were not explicitly speaking to moral licensing research; however, their findings are in line with this research area and warrant further investigation of moral licensing at the organizational level. Although research has yet to examine moral licensing at the organizational level, we propose that, because top leaders influence firms’ strategic decision‐making processes (Finkelstein & Hambrick, 1996; Finkelstein, Hambrick, & Cannella, 2009), the implementation of a CSR strategy will license them to engage in organizational wrongdoing. Specifically, given that people desire to appear moral, top leaders may feel pressure to develop socially responsible strategies that consider the needs of multiple stakeholders. The development and implementation of a CSR strategy may help leaders create a moral image of themselves and their firm, and ultimately provide a sense of accumulated moral credits. For instance, prior to the Enron scandal, former CEO Kenneth Lay endowed chaired positions at universities as well as donated vast amounts of money to charity (Philanthropy News Digest, 2001). Indeed, “upon his taking the helm of the corporate merger that resulted in Enron, Lay became a big player in the charitable and philanthropic environment of Houston, the state, and the nation” (Non‐Profit Quarterly, 2006). Such behavior on the part of the leader builds his social responsibility credits, which may license him to commit socially irresponsible behavior in the future. In other words, top leaders may feel that when they have acquired moral credits through a CSR strategy that balances the needs of multiple stakeholders, they can then put forth a strategy that cuts corners or is potentially harmful to stakeholders. A second route that prior CSR could lead to CSiR is through employees’ reactions to their leader's previous CSR strategy. Recall that when a person feels strongly identified with another individual or group, he or she internalizes their past moral behavior and subsequently is more likely to engage in unethical behavior (Kouchaki, 2011). Applied to organizations, employees who identify with their organizational leaders may feel that the prior CSR strategy set by these leaders licenses them to be less careful in their relationship with stakeholders. For example, recall Hayward's implementation of a safety culture at BP. BP employees may have felt that Hayward's actions and more broadly the firm's reputation for being safety conscious carried over to them and they may subsequently have been less careful about monitoring key safety warning signs. In sum, we argue that top leaders are in charge of making key strategic decisions that influence firm outcomes and that when they implement a CSR strategy it helps them accrue moral credits. These moral credits allow them to be less vigilant toward managing stakeholder needs or even engage in unethical behavior without discrediting themselves or their organization. Moreover, employees may feel licensed to engage in unethical behavior toward stakeholders because they have categorized their leader as part of their ingroup and internalized their past CSR action. Formally, we propose: Hypothesis 1: Prior CSR is positively associated with subsequent CSiR.

The Influence of CEO Moral Identity Symbolization on the CSR–CSiR Relationship Although moral licensing research has found that individuals are generally inclined to engage in morally questionable behavior after having engaged in socially desirable behavior, this process runs counter to the fundamental psychological finding that people desire consistency in their beliefs and behaviors (Audia, Locke, & Smith, 2000; Bem, 1972; Festinger, 1957). Thus, recent calls to examine when licensing occurs and whether some people remain consistent in their moral behavior across time have been issued (Merritt, Effron, & Monin, 2010). In other words, it is important to understand when inconsistency trumps people's basic desire for consistency. Some boundary conditions to moral licensing have been suggested, with Mulder and Aquino (in press), for instance, proposing that an individual difference, moral identity, influences the consistency of moral behavior across time. Moral identity is the extent to which being moral is central to one's sense of self (Aquino & Reed, 2002; Blasi, 1984). It is comprised of two components: internalization, which refers to the degree to which moral identity is central to one's sense of self and is privately expressed, and symbolization, which refers to the degree to which moral identity is expressed outwardly to the public through actions and behavior (e.g., Aquino & Reed, 2002; Mayer, Aquino, Greenbaum, & Kuenzi, 2012; Reed & Aquino, 2003). Given our focus on leader moral action (e.g., CSR strategy implementation), rather than the leader's internal moral state, we examine how leader moral identity symbolization affects the relationship between prior CSR and CSiR. Individuals higher on moral identity symbolization engage in activities and behaviors that signal to others that they are moral beings. These individuals may signal their morality through actions because of a genuine desire to communicate the importance of their moral identity. Alternatively, they may engage in moral behaviors because of a self‐interested desire to present themselves in the most positive light as well as manage others’ impressions of them (Bolino, 1999; Goffman, 1959; Winterich, Aquino, Mittal, & Swartz, 2012). In other words, individuals high on moral identity symbolization may engage in moral actions to either highlight their true morality or instead to mislead people about the value they place on morality (i.e., their moral actions may be insincere). Conversely, individuals who are lower on moral identity symbolization make little effort to signal to others that morality is important to them and may even engage in behavior that signals that morality is not valued.1 To formulate our arguments regarding the impact of a leader's moral identity symbolization on the CSR–CSiR relationship, we draw upon Mulder and Aquino's (in press) study that examined the role of moral identity internalization in the relationship between past and subsequent moral action. They found that individuals high on moral identity internalization were less likely to donate money to a charity after previously telling the truth than were individuals low on moral identity internalization who instead remained consistent in their behavior (i.e., they donated money to a charity after previously telling the truth). To explain these seemingly counterintuitive effects, Mulder and Aquino draw on control theory (Carver & Scheier, 1982; Weiner, 1948) to suggest that attaining a moral standard (i.e., a goal of being moral) is more important for individuals high on moral identity internalization than for individuals low on moral identity internalization. When individuals high on moral identity internalization have not met, or, conversely, have surpassed their moral standard, they engage in “discrepancy reduction efforts” (Karoly, 1993, p. 33) to achieve balance between their current and desired moral state. And when individuals high on moral identity internalization have met their moral standard (i.e., satisfied their goal), they focus on other goals, even goals that may result in less ethical behavior. Mulder and Aquino suggest that these efforts to meet a moral standard are what results in either moral cleansing (when individuals high on moral identity internalization have previously engaged in immoral behavior) or moral licensing (when individuals high on moral identity internalization have previously engaged in moral behavior). Although Mulder and Aquino (in press) studied moral identity internalization, their findings provide a foundation for thinking about the role of moral identity symbolization in the CSR–CSiR relationship. We draw upon their consideration of control theory to explore how moral identity symbolization affects the CSR–CSiR relationship. Recall that moral identity symbolization may or may not accurately reflect a person's internalized moral identity. In other words, an individual who is high on moral identity symbolization may engage in moral action simply to present him‐ or herself in a positive light and to manage others’ impressions of them. Therefore, we argue that those high on moral identity symbolization may be more focused on monitoring their behavior to determine if they have accumulated enough moral credits than will those who are low on moral identity symbolization. Further, they will look for signals in their environment that communicate whether they have gleaned enough moral credits to appear moral or not. Given that Mulder and Aquino (in press) find that referencing one's past actions is central to guiding future behavior, one key signal for leaders is prior CSR. First, we consider the case of high levels of prior CSR. High levels of past CSR should signal to leaders with high moral identity symbolization that they have met their moral standard of appearing moral to others. In other words, these leaders will perceive that their past symbolic gestures of morality (i.e., prior CSR) have bolstered their moral credits. As such, this goal of appearing moral will be less accessible or active for leaders high on moral identity symbolization, causing them to engage in fewer symbolic acts that signal their moral image to others. Instead, they may attend to other goals that may have been ignored (e.g., shareholder wealth), potentially resulting in unethical stakeholder management (e.g., harming the environment in an effort to increase profits for shareholders). Alternatively, comparing their current state to a moral standard will not be as important to leaders low on moral identity symbolization. Instead, they are likely to focus on other goals that are more central to their self‐definition. Therefore, leaders low on moral identity symbolization will not be as reactive to past CSR and instead will be more likely to remain consistent in their strategies toward stakeholders, irrespective of past CSR. Next, we consider the case of low levels of prior CSR. In this case, leaders with strong moral identity symbolization will feel that their current state does not meet their standard of appearing moral to others. In other words, they have not accrued enough moral credits to satiate their need to appear moral to themselves and others. Past research finds that when an individual's moral identity is threatened, and thus the desire to meet this moral standard is heightened, he or she will engage in ways to self‐symbolize a positive moral image (Burke, 1991; Skarlicki, Van Jaarsveld, & Walker, 2008). We predict that leaders high on moral identity symbolization may be more likely to attend to this signal and actively work to be viewed as socially responsible by stakeholders. One way in which this may occur is through reduced wrongdoing toward stakeholders. For example, the CEO may publicly communicate to employees that wrongdoing toward stakeholders is unacceptable. On the other hand, as mentioned above, because appearing moral will be less important to leaders low on moral identity symbolization they will engage in the same level of CSiR, irrespective of past CSR. Hypothesis 2: Moral identity symbolization moderates the relationship between prior CSR and CSiR, such that the relationship between CSR and CSiR is more positive when CEOs are high on moral identity symbolization rather than low on moral identity symbolization.

Method Sample We tested our predictions in a business context using a sample consisting of 49 publicly traded Fortune 500 organizations. The organizations examined in this paper were taken from a larger study examining the relationships among CEO characteristics, TMT processes, and firm performance during the time period 1996–2002 (see Wong et al., 2011, for details on the sampling process). Following Wong and her colleagues (2011), because of the qualitative nature of this study, we only examined firms from the 2002 Fortune 500 list that had a reasonable amount of media coverage (i.e., at least 10 articles of approximately 1,000 words of media coverage during our time period). From this process, the list was narrowed to 65 firms. For each of these 65 firms we then collected articles in the business press and books that described the behavior and personality of each firm's CEO. In order to be included in the sample, there needed to be sufficient information on the CEO to be able to complete an assessment using the California Adult Q‐sort, a forced distribution methodology that allows description of a person using first and third party data (CAQ; Block, 1961; Peterson, Smith, Martorana, & Owens, 2003). We used similar guidelines as those used to collect articles for previous at‐a‐distance research (i.e., at least 10 articles of approximately 1,000 words of media coverage during our time period; Peterson et al., 2003; Tetlock, Peterson, McGuire, Chang, & Feld, 1992; Wong et al., 2011). This narrowed our sample to 55 CEOs. Finally, we collected CSR and CSiR data on each firm. Due to missing data, our final sample included 49 organizations. For the organizations that had multiple CEOs during the 1996–2002 time period, we selected those CEOs who were in office during the latter part of our time period.2 The organizations in our sample represented a range of industries (e.g., retail, computer manufacturing) and had an average of $4.0 billion in sales and 138,429 full‐time employees. Examples of organizations in our sample include International Business Machines, Inc., Nike, Inc., and PepsiCo, Inc. Measures Independent variable: CSR Our measure of CSR was obtained from Kinder, Lydenberg, Domini Inc.'s (KLD) social ratings, which is a commonly used measure of CSR that is considered to be a comprehensive measure of multiple stakeholders’ needs (e.g., Agle, Mitchell, & Sonnefeld, 1999; Kotchen & Moon, 2011; Strike et al., 2006). KLD ratings examine both positive action (i.e., strengths) and negative action (i.e., concerns) toward a variety of stakeholders. It should be noted that CSR is often measured as an aggregate of the KLD scores that subtracts KLD concerns from KLD strengths (Graves & Waddock, 1994; Hillman & Keim, 2001; Thomas & Simerly, 1995). However, recent research has suggested that KLD strengths and KLD concerns should be examined separately (e.g., Kacperczyk, 2009; Mattingly & Berman, 2006; Strike et al., 2006). Following Strike and colleagues (2006), we utilize KLD strengths to measure CSR and KLD concerns to measure CSiR. The KLD strength ratings are generated by ascertaining a firm's strengths along seven key stakeholder benchmarks (community relations, diversity, employee relations, environment, product, corporate governance, and human rights). Each benchmark is composed of multiple sub‐indicators. For each of these sub‐indicators, organizations are assigned a dichotomous rating of 1 or 0 for the presence of a CSR strength. For example, for the employee relations benchmark, a CSR strength is when the organization has a history of strong union relations. For the community relations benchmark, an example of a CSR strength is the organization consistently gives more than 1.5% of its net earnings to charity. To obtain our measure of CSR we averaged each firm's KLD strengths scores for the seven benchmarks (Deckop, Merriman, & Gupta, 2006; Wong et al., 2011). Past research on CSR has found significant industry effects (e.g., Waddock & Graves, 1997), and therefore, following Agle et al. (1999) and Waldman, Siegel, and Javidan (2006), we controlled for firm industry by subtracting the industry mean from the CSR score for each firm. In this study, we used KLD strengths data from 2001 to 2002 to create our CSR measure. Following Agle and his colleagues (1999), we smoothed the data by averaging across the 2 years. The average CSR score adjusted for industry was .48 (SD = .43). The higher the CSR score the more positive actions a firm engaged in toward its seven stakeholders. Independent variable: Moral identity symbolization The most common method utilized to assess moral identity is self‐report measures of the importance of moral traits in terms of centrality to self‐definition (internalization) as well as the extent to which one's behaviors emblemize these traits (symbolization; Aquino & Reed, 2002). Aquino and Reed (2002) developed a list of nine traits that describe a moral person: caring, compassionate, fair, friendly, generous, helpful, hardworking, honest, and kind. This list of traits is not meant to be exhaustive but rather representative of a moral person. In their study, participants were asked to imagine how a person who represented these traits would think, feel, and act and then respond to five items measuring internalization (e.g., “Being someone who has these characteristics is an important part of who I am”) and five items measuring symbolization through the behaviors, activities, and organizational memberships that reflected these traits (e.g., “I am actively involved in activities that communicate to others that I have these characteristics”). They correlated these self‐report scales with other methods of moral identity assessment including judges’ ratings of morality based on individuals’ descriptions of their personality and found for example that self‐ratings on moral identity symbolization were positively correlated with third party morality ratings (Study 4). Thus, Aquino and Reed not only identify moral identity traits but also indicate that third party observers can reliably assess moral identity symbolization. With regard to our study, because top leaders typically do not submit to lengthy psychological tests (Hambrick, 2007; Sumanth & Cable, 2011), we adapted the measurement method Aquino and Reed (2002) utilized in their fourth study by obtaining judges’ ratings of leaders’ moral identity symbolization based on publicly available information on their personal characteristics (i.e., behavior and personality). Specifically, we employed a Q‐sort methodology. The Q‐sort methodology asks evaluators to observe or read information on a person, group, or organization and then rank a set of statements or items in terms of how characteristic they are of the person (i.e., California Adult Q‐sort; Block, 1961), group (Group Dynamics Q‐sort; Peterson, Owens, Tetlock, Fan, & Martorana, 1998), or organizational culture (Organizational Culture Q‐sort, O'Reilly, Chatman, & Caldwell, 1991). One of the strengths of the Q‐sort methodology is that it combines “the descriptive richness of the qualitative approach with the rigor of a quantitative approach by creating a common data language” (Peterson, Owens, & Martorana, 1999, p. 107) that allows for comparisons among individuals, groups, or organizations across time and situations. Moreover, it is possible to utilize publicly available information to conduct the Q‐sort, thus allowing measurement of individuals or groups that are normally difficult to examine (Peterson et al., 1999). Because we were interested in examining individual leaders, we utilized the California Adult Q‐sort (CAQ; Block, 1961), which is composed of 100 items that describe an individual. Sample items from the CAQ include “Behaves in an assertive fashion; not afraid to express opinions; speaks up to get what s/he wants” and “Has a rapid personal tempo; behaves and acts quickly; is fast‐paced.” Research assistants, who were blind to the study's hypotheses, independently reviewed publicly available interviews and biographies of each CEO in our sample and then rated him/her in terms of each of the 100 items by sorting them into those that were most characteristic and least characteristic of the CEO. Our measure of moral identity symbolization was composed of seven CAQ items that were representative of moral traits (Aquino & Reed, 2002). The seven items of our moral identity symbolization scale were “Is dependable and responsible (CAQ item 2),” “Is giving, generous toward others (regardless of the motivation) (CAQ item 5),” “Behaves in a sympathetic and considerate manner (CAQ item 17),” “Is productive, gets things done (CAQ item 26),” “Has warmth; has capacity for close relationships, compassionate (CAQ item 35),” “Makes moral judgments; judges self and others in terms of right and wrong (regardless of the nature of the moral code, whether traditional or liberal) (CAQ item 41),” and “Behaves ethically; has a personal value system and is faithful to it (CAQ item 70).” These items overlap with Aquino and Reed's (2002) traits of caring, compassionate, kind, friendly, generous, helpful, hardworking, honest, and fair. For instance, Aquino and Reed's traits caring, compassionate, kind, friendly, and helpful relate to the CAQ items that focus on warmth, capacity for close relationships, sympathetic, considerate, and giving. Aquino and Reed's trait hardworking relates to the CAQ items that focus on being dependable and responsible, and being productive. Finally, Aquino and Reed's traits fair and honest can be seen in the CAQ items that mention morality and ethical behavior. The mean moral identity symbolization score was 6.68 (SD = .75) and was highly reliable (α = .81). The specific procedure for the CAQ is as follows. We first collected qualitative information on each CEO in our sample from books as well as the business press. To compile these information packets we searched for interviews, speeches, biographies, and autobiographies in the business press (e.g., Business Week, Fortune, and Forbes) and books that discussed the CEO's background, behaviors and personality. Each packet included biographies on the CEO from Marquis Biographies and the International Directory of Business Biographies, as well as multiple articles or interviews from the varied sources mentioned above. Articles or book chapters had to be at least 1,000 words in length to be included in the packet. Packets (i.e., total pages of articles) were required to be at a minimum 20 pages and at a maximum 50 pages (procedure adapted from Peterson et al., 2003). Research assistants were trained in a series of sessions. They were asked to read through a CEO's packet and write in the margins or mark areas that seemed of particular significance to the CAQ items. After reading through the packet, research assistants were asked to describe the CEO along the 100 CAQ items. Research assistants were told to rely on the information from the packet when making their assessments and disregard any prior knowledge or preconceived notions of the CEO. The specific procedure is as follows: Research assistants began the process by sorting the cards into three stacks, starting with Item 1. Each of the 100 CAQ items is presented on a 2×3 card, which was obtained from Mind Garden, Inc. In the first stack, coders placed those cards for which the qualities or traits were least characteristic of the CEO. In the second stack, research assistants placed all cards for which the qualities or traits were moderately characteristic of the CEO. And finally, the third stack included all cards for which the qualities or traits were most characteristic of the CEO. For this initial card sort, the research assistants were told that no attention needed to be paid to the number of cards falling into each grouping. Once the three stacks had been established, the coders were then told to further differentiate the items, or cards, into more categories, placing the least characteristic statements at one end and the most characteristic statements at the other until the items were rank ordered into one of nine categories that ranged from 1 = most uncharacteristic to 9 = most characteristic. Because the CAQ follows a forced distribution, research assistants were told to place fewer cards in the extremes (i.e., only five CAQ items can be rated a 1 and only five CAQ items can be rated 9) and more cards in the center categories (i.e., 18 CAQ items can be rated a 5). This process is advantageous compared to standard scale‐based measurements because it increases interrater reliability as well as the predictive power of the measure. At least two research assistants independently reviewed a CEO's qualitative packet and then individually rated the CEO on the 100 CAQ items. Our interrater reliability on our sample of 49 organizations ranged from α = .71 to α = .97 with an average of α = .84 (SD = .06). Thus, the research assistants’ scores were averaged to create one CAQ sort per CEO. Dependent variable: CSiR As with our measure of CSR, our measure of CSiR was obtained from KLD's CSR ratings, specifically the KLD concerns. The KLD concerns ratings are generated by ascertaining a firm's concerning behavior in relation to the same seven key stakeholders (community relations, diversity, employee relations, environment, product, corporate governance, and human rights), with each stakeholder benchmark composed of several sub‐indicators. For each of the stakeholder benchmark sub‐indicators, organizations are assigned a dichotomous rating of 1 or 0 for the presence of a CSR concern. For example, for the employee relations benchmark, a CSR concern is when the organization has poor union relations. For the community relations benchmark, an example of a CSR concern is the organization has engaged in controversial behavior that has negatively impacted a community's economics (e.g., reducing property value within a community). Following previous research (e.g., Strike et al., 2006), we averaged the KLD concerns scores along the seven key stakeholder benchmarks to measure CSiR. In addition, we controlled for firm industry by subtracting the industry mean from the CSiR score for each firm (Agle et al., 1999; Waldman, Siegel, & Javidan, 2006; Wong et al., 2011). In this study, our dependent variable was comprised of KLD concern data from 2003 to 2004 in order to cover the 2 years after the measure of our independent variable (i.e., CSR). However, additional analysis of time lags ranging between 1 and 4 years for our dependent variable indicate the same pattern of results. Following Agle and his colleagues (1999), we smoothed the data by averaging across the 2 years (i.e., 2003 and 2004). The average social irresponsibility score adjusted for industry was .50 (SD = .44). The higher the CSiR score, the more a firm engaged in negative actions toward the seven stakeholders. Control variables Previous research on social responsibility has found that it is related to firm financial performance (e.g., return on assets; see Margolis & Walsh, 2003 for a review). Thus, we control for industry‐adjusted firm financial performance using firm return on assets (Agle et al., 1999) averaged over the sample time period. In addition, following David and colleagues (2007) and Waldman, Siegel, and Javidan (2006), we control for industry‐adjusted CSiR over the sample time period. Researchers have also noted the possible effects of firm size, and therefore we control for firm size (log of the average sales over the sample time period; e.g., Finkelstein & Hambrick, 1990; Graves & Waddock, 1994; Waddock & Graves, 1997). Following Waldman, Siegel, and Javidan (2006), we also control for CEO tenure (logged). Finally, because our CAQ packets varied in the number of pages, we control for packet length (in pages, logged). Industry data were collected from Dun & Bradstreet's Industry Norms and Key Business Ratios, firm data were obtained from COMPUSTAT, and data for CEO tenure were obtained from Dun and Bradstreet's Book of Corporate Managements.

Results Table 1 presents the means, standard deviations, and correlations among the study variables. Before analyzing our data, we examined the variance inflation factor (VIF) for all variables in our model to check for multicollinearity. The highest observed VIF score in our study variables was 1.61, which suggests that multicollinearity was not a problem as this value is well below the conventional cutoff of 10.00 (Neter, Wasserman, & Kutner, 1989). Table 1. Descriptive Statistics and Correlations Variables Mean SD 1 2 3 4 5 6 7 1. CAQ packet length (in packet length) 3.61 .31 2. CEO tenure (in tenure) 2.84 .90 −.19 3. Firm size (in sales) 23.97 .96 .14 .08 4. Prior return on sales, industry adjusted (1996–2002) 1.66 5.93 −.09 .38* −.10 5. Prior CSiR, industry adjusted (1996–2002) .27 .38 .12 .02 .59** –.15 6. Prior CSR, industry adjusted (2001–2002) .48 .43 .30* −.13 .29* −.15 .31* 7. CEO moral identity symbolization 6.68 .75 −.11 .02 −.11 −.07 −.06 .11 8. CSiR, industry adjusted (2003–2004) .50 .44 .09 .08 .61** −.15 .89** .42** −.01 To test our hypotheses, we regressed CSiR on the control variables, prior CSR, CEO moral identity symbolization, and their interaction term. Table 2 shows the results of these regressions. In Model 1, the base model, we regressed CSiR on only our control variables: length of the CAQ packet, CEO tenure, firm size, prior firm financial performance, and prior CSiR. The overall model results indicate that the control variables explain approximately 80% of the variance in CSiR. Table 2. Hierarchical Regression Results for Corporate Social Irresponsibility on Prior Corporate Social Responsibility and CEO Moral Identity Symbolization Dependent variable: CSiR (2003–2004) Variables Model 1: Base model Model 2: Main effect model Model 3: Full model with interaction Step 1: Control variables CAQ packet length (in packet length) −.02 −.08 −.09 CEO tenure (in tenure) .03 .04 .04 Firm size (in sales) .06 .05 .07 Prior return on sales (1996–2002) −.00 −.00 −.00 Prior CSiR (1996–2002) .92** .88** .83** Step 2: Main effects Prior CSR (2001–2002) .18* −1.34† CEO moral identity symbolization .01 −.08 Step 3: Interaction Prior CSR × CEO moral identity symbolization .22* Overall model R2 .80 .83 .85 Adjusted R2 .78 .80 .82 Change in R2 .03 .02 F statistic for change 3.18* 4.56* Overall F statistic 34.85** 28.33** 27.50** In Model 2, we regressed CSiR on the control variables, prior CSR and CEO moral identity symbolization. As seen in Step 2 of Model 2, prior CSR has a significant positive relationship with CSiR (ß = .18, p < .05), providing support for Hypothesis 1. Moral identity symbolization has no relationship with CSiR (ß = .01, ns). The overall model results indicate that the control variables, prior CSR, and moral identity symbolization explain approximately 83% of the variance in CSiR (p < .05). In Model 3 we test the prediction that CEO moral identity symbolization moderates the relationship between prior CSR and CSiR such that the relationship between CSR and CSiR is more positive when CEOs are high on moral identity symbolization rather than low on moral identity symbolization (Hypothesis 2). In support of this hypothesis, the interaction between prior CSR and moral identity symbolization is significant (ß = .22, p < .05). Simple slope analyses (Aiken & West, 1991) provide support for our hypothesis by showing that CSiR is positively related to prior CSR among CEOs high on moral identity symbolization (b = .43, SE = .14, t = 3.13, p < .01), whereas CSiR is unrelated to prior CSR among CEOs low on moral identity symbolization (b = –.31, SE = .24, t = –1.29, ns). These simple effects are depicted in Figure 1. This step explains 2% of the incremental variance in CSiR and yields an R2 of .85 (p < .05). Figure 1 Open in figure viewerPowerPoint Moderating Role of CEO Moral Identity Symbolization on Prior Corporate Social Responsibility and Corporate Social Irresponsibility Relationship. Additional Analyses Although our analyses indicate support for our prediction that CSR is positively associated with future CSiR (Hypothesis 1), because previous research has found that CSiR leads to CSR (Kotchen & Moon, 2011), we also tested our data for reverse causality (i.e., CSiR associating with subsequent CSR). We find that past CSiR (2001–2002) is positively related to CSR (2003–2004; r = .33, p < .01). However, when we conduct regression analyses to examine how past CSiR relates to subsequent CSR and control for firm size, past firm financial performance, and past CSR, we do not find support for the reverse causal order (ß = .15, ns) nor do we find support for the interaction between past CSiR and CEO moral identity symbolization predicting subsequent CSR (ß = .03, ns). These results support the theoretical framework upon which we draw (i.e., moral licensing) and provide greater confidence in our finding that CSR predicts CSiR and that CEO moral identity symbolization moderates this relationship.

Discussion At its height, Enron was Houston's go‐to corporate philanthropist. The company gave heavily to arts groups, scholarship funds, and local medical centers (Philanthropy News Digest, 2001). But for all this good, it also engaged in surreptitious accounting fraud that eventually cost shareholders $11 billion when its stock price dropped from nearly $100 per share to $1 at the end of 2001. Such corporate misconduct toward stakeholders continues to be a common occurrence. In response, business practitioners and academics alike have considered how to increase CSR and, specifically, how leaders can facilitate CSR. This research has been instrumental to our understanding of how to encourage CSR yet it does not directly speak to the impetus for much of this research: Whether and when do firms behave badly? In this paper, we have begun to address these questions by considering how leaders’ psychological processes and personal characteristics relate to CSiR. Using archival data on Fortune 500 firms, we demonstrated that firms that engaged in prior socially responsible behavior are more likely to then engage in socially irresponsible behavior. In addition, we found that CEO moral identity symbolization moderates the relationship between prior CSR and CSiR, such that the relationship is stronger for CEOs who are high on moral identity symbolization rather than low on moral identity symbolization. Theoretical Contributions This study has several theoretical contributions for research on CSR, strategic leadership, moral licensing, and moral identity. First, our work contributes to research on CSR in a number of ways. Past research has assumed that engaging in CSR has positive consequences and, thus, has focused on how CSR leads to positive outcomes such as firm financial performance. However, given that much psychological research has shown that good behavior leads to bad behavior (e.g., psychological licensing), it is important to understand whether CSR also results in detrimental outcomes. Our research draws on this psychological research to find that CSR can indeed result in negative consequences: heightened CSiR. We also contribute to CSR research in that previous research has been limited in its examination of the antecedents and consequences of CSiR (Campbell, 2007; Greenwood, 2007). To the extent that it has been examined, previous CSR researchers have focused on how prior CSiR affects subsequent CSR (e.g., Chatterji & Toffel, 2010; Muller & Kräussl, 2011). We consider the opposite relationship—that of how CSR affects subsequent CSiR. In addition, the research that has been conducted on CSiR has been at the firm‐level only. Our research complements this research by taking a microlevel approach and exploring the role of strategic leaders in shaping the CSR–CSiR relationship. To better understand this relationship, we draw on strategic leadership and moral licensing research to examine how a leader's implementation of past CSR might license them to subsequently be less ethical and careful when managing stakeholders’ needs. As such, we contribute to our understanding of CSR and CSiR in companies such as Enron or BP. Related to this contribution, researchers have called for studies that utilize social psychological theories to understand CSR, but little research has been conducted applying this lens (e.g., Rupp et al., 2011). Our research responds to this call by examining how moral identity relates to CSiR. We found that moral identity symbolization was not related to CSiR. At first blush this finding appears to run counter to moral identity research that finds moral identity to be positively related to prosocial behavior and negatively associated with antisocial behavior (see Shao, Aquino, & Freeman, 2008 for a review). However, recent research suggests the importance of contextual factors, such as past behavior, in understanding moral identity's effects on moral outcomes (Mulder & Aquino, in press). In line with this research, we find that prior CSR is an important contextual factor that influences whether leaders’ moral identity symbolization affects CSiR. This examination helps us better understand leaders’ attention to CSR and how it affects subsequent strategic choices about CSiR. Second, we contribute to strategic leadership research in which scholars have repeatedly called for studies that consider how the psychological processes of top leaders influence firm‐level outcomes (Cannella & Monroe, 1997; Hambrick, 2007). Researchers have noted the difficulty of conducting this research because top leaders generally do not submit themselves to psychological assessments (Cannella & Monroe, 1997; Hambrick, 2007; Hermann, 1999; Sumanth & Cable, 2011). Although past research has utilized demographic variables as proxies of leaders’ personal characteristics, we contribute to strategic leadership literature by examining an important, but underexamined, personal characteristic of CEOs (i.e., moral identity symbolization) using a novel, unobtrusive methodology, the CAQ. Because the CAQ methodology transforms qualitative information into quantitative data, it allowed us to make comparisons across more CEOs and raters than the case study methodology allows. Third, our study contributes to research on moral licensing. To our knowledge, our study is the first to examine moral licensing at the organizational level using data with high ecological validity, and in this way our research contributes to the burgeoning literature on psychological licensing. Prior moral licensing research has been conducted at the individual level and within experimental settings to show that participants are more likely to engage in morally questionable behavior after having previously engaged in socially desirable behavior. By using data from Fortune 500 companies, we extend theoretical understanding of this psychological process to leaders and show its organizational‐level implications. In addition, we respond to recent calls by researchers to consider how individual differences affect moral licensing (Jordan, Mullen, & Murnighan, 2011; Miller & Effron, 2010). Our findings demonstrate a boundary condition to moral licensing research by showing that leaders who are less focused on their moral actions (i.e., low moral identity symbolization) are more likely to remain consistent. Alternatively, our study suggests that leaders who are high on moral identity symbolization may be more morally aware, as opposed to consistent, and thus know when they can and cannot engage in less ethical behavior without compromising their moral integrity. Therefore, this study contributes to our understanding of when inconsistency trumps people's basic desire for consistency (Bem, 1972; Festinger, 1957). Fourth, we contribute to moral identity research. The majority of research on moral identity has focused on moral identity internalization (see Shao et al., 2008, for a review; Winterich et al., 2012). Few studies have examined the effects of moral identity symbolization, despite research showing that symbolization has significant effects on behavior (e.g., Mayer et al., 2012). As such, one of our theoretical contributions to this research is our examination of symbolization effects. In addition, we contribute to the moral identity research by identifying an alternative method of measurement. Typically, studies have utilized Aquino and Reed's (2002) self‐report measure of moral identity symbolization. However, one potential drawback of this method is that it relies on individuals to report their behavior and their perceptions of themselves. In this way we provide a methodological contribution to this literature by using the Q‐sort methodology in which multiple judges, who were blind to the study hypotheses, rated the leaders on a wide range of behavior resulting in less biased ratings of moral actions. Practical Implications In addition to the theoretical contributions of this paper, our results have practical implications as well. First, our findings suggest that governance structures should be put in place that explicitly hold leaders accountable for both stakeholder management and stakeholder mismanagement. Given leaders’ propensity to engage in wrongdoing after they have previously engaged in CSR, boards should be particularly careful after the successful implementation of a CSR strategy to ensure that leaders do not become complacent. Moreover, boards should set clear rules about intolerance toward unethical or illegal behavior. Boards need to clearly communicate to top leaders that unethical behavior toward stakeholders will not be tolerated and specify the consequences of such mismanagement. Interestingly, Zhong and colleagues (2010) note that highlighting the importance of “doing good” and rewarding ethical behavior may inadvertently cause employees to engage in unethical behavior. As such, boards and leaders should take great caution in considering how to manage the CSR–CSiR relationship. Contrary to what one might expect, our research suggests that leaders who attempt to put forth a moral image are in fact more likely to engage in moral licensing behavior than are those who do not engage in symbolic moral action. Although boards may believe that CEOs who are high on moral identity symbolization can be trusted to behave ethically on a consistent basis, our research instead shows that they may be less likely to remain consistent. In fact, these leaders may be more in tune with managing others’ moral impressions of them and in maintaining a balanced moral scorecard, which may not always result in beneficial outcomes for the firm. This focus may be advantageous when companies have previously engaged in wrongdoing. However, it may be harmful when companies have previously engaged in CSR. Thus, boards should remain particularly vigilant of leaders who put forth a moral image. Limitations and Future Research The theoretical contributions and practical implications of our study should be considered with regard to the study's limitations. First, in this research we attend to moral licensing as one means to understand the short‐term CSR–CSiR relationship. We found robust support for this relationship by first demonstrating that prior CSR predicts future CSiR and second by conducting additional analyses that ruled out reverse causality. Although these analyses provide confidence for observed moral licensing in our present research, other research has found evidence for moral cleansing, whereby morally questionable actions give rise to ethical or prosocial behavior (Carlsmith & Gross, 1969; Sachdeva et al., 2009; Zhong & Liljenquest, 2006). Jordan and colleagues (2011) suggest that moral behavior is dynamic and acts much like a pendulum that swings between moral cleansing and moral licensing. One limitation then of our study is that we only examine a short time period, which may only capture a snapshot of a larger moral licensing‐cleansing cycle. Future research should investigate this pendulum by examining CSR and CSiR over a longer time frame. Second, one might note that the effect sizes in this study are small. Though categorized as small, the effect sizes are comparable with other studies examining CEOs’ impact on CSR (e.g., Deckop et al., 2006; McGuire, Dow, & Argheyd, 2003). In addition, despite the stability of CSiR over time (e.g., the correlation between CSiR 2001–2002 and CSiR 2003–2004 is highly significant), our results indicate that leadership still plays an important role in shaping organizations’ CSR strategies. Finally, small effect sizes can still be impactful when they show a dependent variable (in this case, CSiR) is influenced by an unlikely independent variable (in this case, CSR; Prentice & Miller, 1992). Specifically, Prentice and Miller (1992) note that “showing that an effect holds even under the most unlikely circumstances possible can be as impressive as (or, in some cases, perhaps even more impressive than) showing that it accounts for a great deal of variance” (p. 163). In our case, finding that socially responsible behavior relates to subsequent irresponsible behavior is important. Finally, although the CAQ methodology offers great benefits to examining top leaders at a distance, it is not without its limitations. First, because of the qualitative nature of the CAQ, it is possible that the data were biased because of journalists’ or coders’ biases. Although our use of varied sources to create our packets of information on each CEO as well as our use of multiple coders minimizes the impact of any individual bias, we cannot fully isolate personal prejudgment or bias in our data. Second, our study is restricted in its sample. Given the level of detail required to complete a CAQ, our sample was limited to CEOs of large, American firms that received extensive coverage in the business press. However, given that leaders have more discretion in smaller firms than they do in larger firms (Miller, Ket de Vries, & Toulouse, 1982), it is likely that this sample serves as a conservative test of our study hypotheses (i.e., results might be even stronger within smaller organizations in which leaders have more control of communication and the decision‐making process). Our study suggests a number of future research directions. In our study we focused on an aggregated measure of CSR, arguing that leaders who develop CSR strategies that balance the needs of a variety of stakeholders subsequently are more likely to engage in CSiR that hurts a variety of stakeholders. This general approach is important because it is the first test of moral licensing in relation to CSR strategies. However, it does not disentangle whether leaders are engaging in domain or nondomain specific moral licensing. Following from earlier moral licensing work in experimental settings, which found that moral licensing at the individual‐level can be both domain and nondomain specific (e.g., Mazar & Zhong, 2011; Monin & Miller, 2001), future research should examine whether moral licensing in firms is domain specific (e.g., does positive action toward the environment then lead to negative action toward the environment?) or nondomain specific (e.g., does positive action toward the environment subsequently lead to negative action toward employees?). Another avenue for future research on CSR–CSiR is to explore the moderating effects of other psychological and contextual factors on CSiR. In this paper, we focused on one important psychological characteristic influencing the CSR–CSiR relationship, moral identity symbolization, but other psychological and contextual factors likely exist. For example, factors such as leader cognitive moral development and moral philosophies are related to ethical leadership (see Kish‐Gephart, Harrison, & Treviño, 2010 for a review) and may influence how leaders use past moral behavior as a guide to determine subsequent moral behavior. It is also possible that leaders who are more actively engaged in developing and implementing strategies may simultaneously engage in both CSR and CSiR. Thus, leader personality attributes related to assertiveness and conscientiousness may also be important to consider. With regard to contextual factors, future research should explore how contextual factors such as organizational culture influence the CSR–CSiR relationship. Given that previous research shows paradoxical negative effects of a meritocratic culture on sexism in the workplace (Castilla & Benard, 2010), we may find that organizational cultures intended to increase the ethical standards of the organization actually lead to subsequent CSiR. Future research studies exploring these psychological and contextual factors are important as they will allow us to better understand how leaders can manage both CSR and CSiR. Our findings also have implications for future research on moral licensing. This study, to our knowledge, is the first to demonstrate moral licensing effects at the organizational level in an organizational context, and our interaction results further indicate leader moral regulation is an important determinant of CSiR. However, like much prior moral licensing research, we did not directly test the underlying moral licensing processes as our initial goal was to demonstrate moral licensing at the organizational level. Having done so, our findings warrant future exploration and identification of the mechanisms driving moral licensing effects (e.g., moral credits). For instance, future research could empirically test whether leaders’ moral credits are indeed bolstered after implementing a CSR strategy. To assess the relationship between the firm's CSR ratings and leaders’ moral credits, a future study could code the manner in which CEOs discuss prior CSR through interviews after the release of the KLD ratings. Moreover, future moral licensing research should explore other factors that may influence the moral licensing effect, such as leaders’ identification with their organization. Given that some leaders are promoted from within their firm whereas others are outsiders, their identification level may vary, which in turn may influence the degree to which they internalize moral credits accrued from their firm's past CSR. Relatedly, longer‐tenured CEOs, who helped implement prior CSR strategies, may be more likely to engage in moral licensing than shorter‐tenured CEOs, who did not implement prior CSR strategies. As such, future research should explore how leader identification and tenure, as well as whether a leader was an internal or external hire, influences the CSR–CSiR relationship. Our research also warrants further investigation in the area of moral identity, and in particular, moral identity symbolization. Specifically, future research should pinpoint why leaders who are high on moral identity symbolization are less likely to be consistent in their moral actions than are those who are low on moral identity symbolization. In contrast to our findings, Zhong and colleagues (2010) suggested that individuals with high moral maturity, or people who consider morality as central to their identity, should be more consistent in their moral action over time than those with low moral maturity because it is more critical to them that they maintain a positive moral image. Our research contradicts their prediction by finding that those with high moral identity symbolization are actually less consistent in their moral action. Although our data cannot directly speak to why high moral identity leaders were more likely to engage in moral licensing, it may be that individuals who are high on moral identity symbolization do not strongly identify with being a moral person (i.e., moral identity internalization). Therefore, their engagement in moral licensing may reflect their strategic use of moral actions. Future research should explore the underlying mechanisms linking moral identity symbolization to moral licensing. Future research regarding moral identity measures is also necessary. Moral identity researchers typically use self‐report measures to assess this construct. We relied on third party data to obtain our measure of moral identity symbolization. Using the Q‐sort methodology allowed us to discreetly measure moral identity symbolization; however, questions remain as to the degree of overlap between self‐report measures and the Q‐sort method we utilized. Future research should examine the degree to which self‐report measures of moral identity internalization and symbolization correlate with the Q‐sort measure of symbolization and what implications a mismatch might have on the CSR–CSiR relationship. In addition, future research should examine how this Q‐sort measure of moral identity symbolization correlates with other related constructs such as ethical leadership and a moral personality (Brown & Treviño, 2006; McFerran et al., 2010).

Conclusion Although previous research has examined positive effects of CSR such as how it relates to firm financial performance, we argued that it may also be an antecedent to CSiR. We further argued that it is important to consider how leaders affect CSiR because they guide the strategic decision‐making process for CSR initiatives. We find that CSR is positively related to CSiR and that the relationship is stronger for leaders who are high on moral identity symbolization rather than leaders who are low on moral identity symbolization. These findings provide a critical step in understanding the underlying processes that influence CSiR.

1 Various constructs relate to moral identity symbolization including the aforementioned moral identity internalization (Aquino & Reed, 2002 2006 2010 1999 1959 2012 2012 o, 2006 o, 2006, for a review)” (Mayer et al., 2012

Various constructs relate to moral identity symbolization including the aforementioned moral identity internalization (Aquino & Reed, o, o, 2006, for a review)” (Mayer et al., 2 The CEOs in our sample were in office for the majority of our time period, which was 1996–2002 (i.e., in office for 3.5 years or more). Eleven CEOs out of the 49 CEOs included in our study were in office for fewer than 3.5 years (i.e., nine CEOs were in office for 3 years of the 1996–2002 time period and two CEOs were in office for 2 years of the 1996–2002 time period). These 11 CEOs’ tenures occurred in the latter part of our time period (i.e., 2000–2002) and thus during the independent variable's time period (2001–2002) as well as closer to the dependent variable's time period (2003–2004). When we exclude these 11 CEOs from our analyses, we find the same pattern of results.