CEO ACTIVISM AS A VALUE-BASED CONTROL By Sangmok Lee A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Business Administration – Accounting – Doctor of Philosophy 2024 ABSTRACT CEO activism—CEOs publicly expressing their opinions on social or political issues not directly related to their firms' core operations—has become increasingly prevalent. This study posits that CEO activism functions as an informal control mechanism. Prior research on management control systems suggests that managers use values and belief systems to influence employees’ behavior (“value-based control”). However, CEOs may seek additional means to communicate specific values because formally stated values tend to be broad and less specific by design. Using novel datasets, I find that CEO activism has short-term negative effects on employee sentiment on the firm’s beliefs and values. Moreover, it amplifies the sensitivity of employee turnover to employee sentiment. In the longer run, CEO activism is positively associated with the firm’s operating performance. Further analysis shows that the inflow of better-educated employees increases after CEO activism. Collectively, these findings suggest that CEO activism promotes employee sorting based on value alignment, ultimately shaping workforce composition and firm performance. Copyright by SANGMOK LEE 2024 To Grandma, who would have been proud. To Mom and Dad, who gave me life and education. To Jiyeon and Junu, who make me persevere. iv ACKNOWLEDGEMENTS This dissertation marks the end of a five-year odyssey that has been at once fast and slow, painful and joyful, humbling and enlightening. I am deeply grateful to my dissertation committee for their invaluable guidance and unwavering support: Kyonghee Kim (Chair), Ranjani Krishnan, Mario Schabus, and Jeff Wooldridge. Their passion, insights, and work ethic continue to inspire me. The Department of Accounting and Information Systsems has been essential in sustaining my living and enabling this dissertation. Ken Bills and Kevin Markle have truly demonstrated their care and support as the Ph.D. program directors. Throughout my doctoral journey, I have been fortunate to receive immense support from countless mentors and friends. I thank Bok Baik for introducing me to this fascinating profession. My gratitude extends to Kwangjin Lee and Patrick Woong Ryu, who made me seriously consider this career path. Martin Holzhacker and Adam Presslee taught me the tremendous value of hard work and positive thinking. I could not have asked for better comrades than my fellow Ph.D. students. I reserve the deepest gratitude for my family. My grandmother, Seja Ki, would have been proud to witness this achievement. I am indebted to my parents, Jihwa Lee and Seokhyo Choi, for nurturing my curiosity and determination. I thank my wife, Jiyeon Kim, for her patience with the chaos within me. Finally, I am grateful to my son Junu for his presence. I would not have persevered without him. v TABLE OF CONTENTS I. INTRODUCTION ....................................................................................................................... 1 II. RELATED LITERATURE AND HYPOTHESIS DEVELOPMENT ....................................... 9 III. RESEARCH DESIGN ............................................................................................................ 16 IV. RESULTS ................................................................................................................................ 22 V. ADDITIONAL ANALYSES .................................................................................................... 30 VI. CONCLUSION....................................................................................................................... 33 REFERENCES ............................................................................................................................. 35 APPENDIX A: VARIABLE DEFINITIONS ............................................................................... 41 APPENDIX B: EXAMPLES OF CEO ACTIVISM ..................................................................... 44 APPENDIX C: CATEGORIES OF ACTIVISM BY TOPIC ....................................................... 45 APPENDIX D: EXAMPLE OF IDENTIFICATION IN THE FIRM-MONTH PANEL ............. 46 APPENDIX E: DYNAMIC ANALYSES ..................................................................................... 47 vi I. INTRODUCTION In January 2017, Starbucks’ then-CEO Howard Schultz issued a statement to the company’s employees noting that “the promise of the American Dream [is] being called into question.” The statement was in response to then-U.S. president Donald Trump’s executive order that temporarily banned citizens of seven African and Middle Eastern countries from entering the U.S. In addition to denouncing the order, Schultz pledged that his company would create jobs for 10,000 refugees over five years as a counteraction. This statement was instantly met with both support and backlash, as some welcomed the company being transparent about a socially contentious issue and others taunted Starbucks for stepping into a sociopolitical debate (Vaughan and Rushe 2017). An increasing number of CEOs are publicly expressing their stances on social or political issues that are not directly related to their firms’ core businesses, a practice referred to as CEO activism. As can be seen from the above anecdote, however, CEO activism can be a “double-edged sword” (Larcker et al. 2018): it is often subject to strongly polarized responses, which make the overall responses from stakeholders and the public difficult to predict. Given the galvanizing nature of sociopolitical issues and uncertain responses from stakeholders, why do CEOs weigh in on contentious issues in public domains? I posit that CEO activism functions as an informal control mechanism, facilitating employee sorting based on value alignment. Prior studies on management control systems document that managers use values and belief systems to control employees’ behavior (“value-based control”) (Simons 1995; Malmi and Brown 2008). Firms have missions, credos, visions, and other formally stated values promulgated internally and externally. However, formal values tend to be broad and are limited in conveying contextual information about the alignment of values and beliefs between the firm and its employees (Bartkus and Glassman 2008). Managers, therefore, look for 1 supplementary mechanisms to saliently communicate the values to employees (Kraus, Kennergren, and von Unge 2017). CEO activism often ensues after sociopolitical events and is conducive to reinforcing existing credos and building employee trust in a nuanced way. Prior studies suggest that CEO activism can be a tool for aligning values with stakeholders (Chatterji and Toffel 2018; Hambrick and Wowak 2021; Mkrtchyan, Sandvik, and Zhu 2023). For example, several works regard CEO activism as a manager’s strategic decision to appeal to consumer segments with a certain political leaning (Hou and Poliquin 2023; Jin et al. 2024; Homroy and Gangopadhyay 2023). However, less research has explored how managers might use CEO activism to demonstrate value alignment with employees. While some studies examine how employees respond to CEO activism (Burbano 2021; Wowak, Busenbark, and Hambrick 2022), whether the mechanism and implications of these responses are as a part of control system remains understudied. This study addresses that gap by providing large-sample evidence that CEO activism helps managers align values with employees by facilitating employee sorting. Publicly communicated values are essential to signal the management’s commitments (Simons 1994). CEO activism is a form of nonfinancial, qualitative disclosure that reveals the CEO’s sociopolitical stances (Hambrick and Wowak 2021; Chen, Dechow, and Tan 2022). Like other types of corporate disclosures, CEO activism entails the cost-benefit tradeoff. It can subject the CEO or the firm to significant costs, including damage to the CEO’s personal reputation and backlash from discontented stakeholders. Given these associated costs, CEO activism is a costly signal that disciplines the CEO's actions (Stocken 2000; Kartik 2009). A major hurdle in testing whether CEO activism serves as a management control is that management controls operate as a system with interdependent components (Grabner and Moers 2013). Internal workings of interdependent components, if observable at all to external parties, 2 pose measurement challenges and are difficult to discern. Therefore, I conduct a series of analyses, the collective results of which can provide evidence that CEO activism is a component of the firm’s belief system (Simon 1995). Specifically, I examine the effect of CEO activism on employee sentiment, employee turnover, and the firms’ subsequent operating performance to operationalize the research question. My first hypothesis investigates the effects of CEO activism on employee sentiment as captured by employees’ ratings of their firms’ senior leadership and culture, because it reflects employees’ perception of the controls implemented by the management (Tessier and Otley 2012). Prior studies find that individuals feel animosity toward others who have opposing political views or partisan identification (Iyengar et al. 2019). Relatedly, in corporate settings in which employees may agree about certain issues, the literature finds that negative sentiment prevails because individuals hold negative perceptions toward “the other faction” (Dimant 2023) or that the act of revealing a stance demotivates employees who disagree (Burbano 2021). Because CEOs expose themselves to such negative sentiment by engaging in activism, I predict that CEO activism negatively affects employee sentiment. I test the above prediction using S&P 1500 firms from 2011 to 2022. I match these firms with Glassdoor ratings and LinkedIn profiles provided by Revelio Labs, a third-party provider of labor- related data, and data on CEO activism hand-collected from Factiva and X (formerly Twitter). Matching the data across these sources yields a panel of 59,163 firm-month observations. To validate the value-relevance of CEO activism in the short term, I test stock market reactions to CEO activism. I find that relative to those of other firms in the sample (“control firms”), abnormal 3 returns of firms whose CEOs engage in activism (“activism firms”) are lower by 0.111 percentage points on the trading day immediately following the occurrence of CEO activism.1 I examine employees’ Glassdoor ratings of their firms across dimensions that capture their (dis)content toward the firm’s culture and senior management, and their overall assessments of the firm. The results show that for a firm with mean ratings, the ratings decline by 1.54% to 3.28% in the month following CEO activism, relative to those during the pre-CEO activism period within the same firms and at control firms. This result suggests that employees express an economically significant degree of discontent related to the CEO’s activism. I further categorize CEO activism into six topical groups to explore whether CEO activism related to diverse topics has differential effects on employee sentiment. The results reveal that while CEO activism on some topics has mixed or insignificant effects on employee ratings, activism related to politics results in significant declines across three rating dimensions. Next, I investigate the effect of CEO activism on employee turnover. Theories in economics contend that congruence in values and identities is a crucial determinant of individuals’ employment decisions (Akerlof and Kranton 2000). Given the partisan nature of topics covered in CEO activism, I hypothesize that such activism facilitates the sorting of employees, whereby CEO activism appeals to incumbent employees whose beliefs align with the CEO’s stance and alienates those whose beliefs do not. Consistent with this hypothesis, I find that CEO activism more than doubles the sensitivity of employee turnover to employee sentiment during the six-month period following CEO activism. These results suggest that CEO activism significantly intensifies 1 Prior studies analyzing short-window stock market responses to CEO activism present both positive (Mkrtchyan, Sandvik, and Zhu 2023; Brownen-Trinh and Orujov 2023) and negative reactions (Bhagwat et al. 2020; Bedendo and Siming 2021). The mixed empirical evidence indicates that market reactions to CEO activism are sensitive to sample construction, specific return measures, and the nature and content of CEO activism. 4 employee sorting at activism firms. I next decompose employee sentiment into three dimensions that likely affect their employment decisions (i.e., top management, firm culture, and politics) and find that CEO activism has the strongest impact on the ties between employees’ sentiment toward firm culture and politics. The last part of my main tests relates CEO activism to the firm’s operating performance. To the extent that the churning of employees improves value alignments between the firm and employees, I expect CEO activism to be associated with subsequent improvement in employee productivity and operating performance. This is consistent with prior studies that examine performance outcomes of employees’ intrinsic motivation (Besley and Ghatak 2005) or cultural homogeneity (Van den Steen 2010). I find that CEO activism is associated with a 3%-increase (11.4%-increase) in the firm’s year-ahead productivity (profitability). Collectively, the results discussed above suggest that CEO activism is associated with both significant labor-related costs in the short term and improved operating performance in the long term. These findings are consistent with the proposition that CEO activism functions as an informal control, prompting employees whose values are misaligned to voluntarily separate, ultimately fostering a more congruent and productive workforce. I supplement the above findings with additional analyses to rule out alternative explanations. First, the observed decline in employee sentiment following CEO activism could alternatively suggest that employees view the firm unfavorably due to factors unrelated to the firm’s values and beliefs. For instance, employees might be reacting negatively to other firm-level events, and their sentiment on other dimensions could spill over into their assessment of the firm’s culture and senior 5 management.2 To address this concern, I conduct falsification tests in which I find that CEO activism does not significantly impact employees’ perceptions of career opportunities, work-life balance, or the firm’s business outlook—dimensions of employee sentiment unrelated to the firm’s values and beliefs. Second, the decline in employee sentiment followed by the increased employee turnover may imply that firms experience poor employee morale for an extended period following CEO activism. This alternative explanation can also undermine the positive associations I have established between CEO activism and the firm's subsequent productivity and profitability. I test effects of CEO activism on employees’ ratings over three-, six-, or twelve-month windows to clarify this point. I find no significant effects of CEO activism on employee sentiment over any of these windows. The lack of significance suggests that the decline in employee sentiment at firms engaged in CEO activism is temporary. Third, the significant increase in employee turnover post-CEO activism may raise concerns that activism firms are reducing their workforce. I find no significant association between CEO activism and the subsequent number of employees at the firm. Cross-sectional tests also reveal that while the magnitude of employee turnover is consistent across all education levels, the data shows an increased inflow of employees with higher levels of education following CEO activism. This result provides suggestive evidence that as CEO activism facilitates employee sorting, these firms 2 One may also suspect that CEOs opportunistically engage in activism, possibly to deflect stakeholders’ attentions away from poor performance or other negative events. However, descriptive statistics and univariate tests of short-window returns surrounding CEO activism imply that performance in the stock market is unlikely to drive CEO activism. Mkrtchyan, Sandvik, and Zhu (2023) show that the occurrences of CEO activism are not clustered close to or far from earnings announcements, ruling out the suspicion that CEOs strategically time CEO activism to mask bad news. 6 attract high quality employees who may contribute to the subsequent improvements in firm-level operating performance.3 This study contributes to the nascent literature on CEO activism by examining the role of CEO activism in aligning a firm's values with those of its employees. While recent research has investigated employee reactions to CEO activism (Burbano 2021, Wowak et al. 2022, Mkrtchyan, Sandvik, and Xu 2024), this study stands apart in two significant ways. First, from a conceptual standpoint, I bridge CEO activism with the burgeoning literature on managers' use of informal controls. The literature provides theoretical models and empirical evidence on corporate culture as a “soft” control component (Guiso, Sapienza, and Zingales 2015b; Cai 2023; Pacelli, Shi, and Zou 2022), but the use of informally communicated values has received less attention (Guiso, Sapienza, and Zingales 2015a; Kachelmeier et al. 2016). My research fills this gap by illustrating how CEO activism can be integral to value-based control through informal communication channels. Second, on the empirical front, I add to the literature by presenting novel, large-sample evidence that CEO activism subsequently amplifies the sensitivity of employee turnover to employee sentiment and is associated with an increased inflow of better-educated employees. These findings suggest CEO activism is a meaningful channel influencing employees’ employment decisions.4 Furthermore, this research complements a body of work that probes into managers’ nonfinancial voluntary disclosures. A stream of studies in accounting examines how firms’ voluntary disclosures on CSR activities relate to their real actions (Dhaliwal et al. 2011). Given 3 I acknowledge that the employee inflow is a noisy measure of firms’ recruiting. In addition to CEO activism, the firms’ hiring policies and the underlying economy are likely to heavily influence the recruiting of new employees. 4 Relatedly, several studies present evidence that political misalignments with the CEO or other members of the leadership affect turnover of senior executives (Babenko, Fedaseyeu, and Zhang 2020; Kempf, Fos, and Tsoutsoura 2023). 7 that the role of traditional and social media as vehicles for information dissemination is becoming increasingly important (Miller and Skinner 2015; Elliott, Grant, and Hodge 2018), I contribute to the literature by investigating the association between CEO activism and vital firm-level outcomes within a control system. This perspective responds to Van der Stede (2011), which calls for management accounting research that studies how disclosures relate to firms’ “internal businesses.” 8 II. RELATED LITERATURE AND HYPOTHESIS DEVELOPMENT Background of CEO Activism Corporate CEOs have traditionally refrained from openly taking stances on socially debated issues (Chatterji and Toffel 2018; Wowak et al. 2022). However, today’s CEOs are increasingly vocal about such issues as gun control, same-sex marriages, and immigration that are rarely directly related to their businesses.5 The recent surge of CEO activism suggests CEOs are increasingly motivated to take stances on contentious issues, in part reflecting an increased level of political polarization among the general population (Boxell et al. 2017) and particularly, corporate executives (Kempf et al. 2023). Anecdotal and academic evidence suggests that CEO activism is a rational decision to demonstrate value alignments with stakeholders, who have increasingly shown significant demand for CEO activism (e.g., Larcker et al. 2018; Chatterji and Toffel 2019; Hambrick and Wowak 2021). Edelman (2022) reports that 60% of its global respondents “expect the CEO to speak publicly about controversial social and political issues” that they care about, and 81% of the respondents believe CEOs should be “personally visible when discussing public policy with external stakeholders.” Some CEO activism appears to strategically target consumers with a specific political leaning (Hou and Poliquin 2023; Homroy and Gangopadhyay 2023). The literature also documents that CEOs’ personal values, status and other individual attributes, and the tendency to imitate other CEOs drive CEO activism (Gupta and Misangyi 2018; Hambrick and Wowak 2021; Mkrtchyan, Sandvik, and Zhu 2023). 5 The proportion of S&P 500 firms involved in CEO activism was less than 1% in 2011 but has soared to 37% by 2019 (Mkrtchyan, Sandvik, and Zhu 2023). 9 Several early studies in management, marketing, and finance document the consequences of CEO activism. These studies provide mixed evidence of effects of CEO activism on stock market responses (Bhagwat et al. 2020; Bedendo and Siming 2021; Mkrtchyan, Sandvik, and Zhu 2023; Brownen-Trinh and Orujov 2023), consumer reactions (Chatterji and Toffel 2019; Hou and Poliquin 2023, Jin et al. 2024), investor reactions (Durney et al. 2022), and employee reactions (Burbano 2021; Wowak et al. 2022; Mkrtchyan, Sandvik, and Xu 2024). The accounting literature reports empirical evidence on nonfinancial consequences of firms openly supporting or opposing a specific sociopolitical event. Chen et al. (2022) examine S&P 1500 firms disclosing support for Black Lives Matter (BLM) movements on their websites and social media in 2020. The authors find that firms with more inclusive corporate cultures and pressure from stakeholder networks were more likely to disclose support for the BLM movements. Jin et al. (2024) present empirical evidence that firms that took a political stance against Georgia’s voting rights law in 2016 met with reduced customer traffic at their retail stores. CEO activism is distinct from, albeit related to, CSR activities in two dimensions. First, CEO activism mostly involves a CEO taking a stance on polarizing issues (e.g., immigration policies, gun control), whereas CSR activities are typically centered around activities that are less subject to social debates (e.g., initiatives for clean water, healthcare) (Bhagwat et al. 2020). Second, CEO activism events tend to be symbolic communications and do not necessarily incur direct expenses at the firm level (Hambrick and Wowak 2021), while CSR activities entail the use of corporate resources. Similarly, CEO activism also differs from firms’ attempts to exert political influences. Whereas CEO activism occurs in the public domain and is visible, firms tend to be covert about their political activities (e.g., contributions to politicians, lobbying). The corporate political involvements consequently receive less attention from internal and external stakeholders than do 10 CEO activism events. Firms’ political contributions may also reflect a firm-level effort to hedge political risk (Christensen et al. 2022) and may be difficult to attribute precisely to the CEO. These features warrant further investigation of CEO activism as a unique form of communication. CEO Activism as an Informal Control Simons (1995) introduces four “levers of control” that managers use to implement the core business strategies and guide employees’ behavior. Among the levers of control are belief systems, which broadly define values that employees should pursue.6 Firms use mission statements and credos to formalize the beliefs that they want employees to espouse, but the formal values are often criticized for being generic and lacking context (Bartkus and Glassman 2008). In extreme cases, the formally communicated values exist “on paper,” with little salience or relevance to employees. Managers therefore may look for mechanisms to ingrain and communicate the values to employees in context-specific manners.7 Importantly, control components rarely function in isolation and can be formal or informal. Managers commonly deploy informal controls incremental to formal ones. For example, Kachelmeier et al. (2016) present experimental results showing that the presence of a nonbinding, informal value statement functions as an informal input control that incrementally constrains employees. Akinyele et al. (2020) show that salience of such informal statements affects employees’ subsequent performance. Relatedly, Graham et al. (2022) exhibit a framework of corporate institutions in which corporate culture, comprised of values and norms, is an informal institution that complements formal institutions (e.g., human resources and governance). 6 Simons (1995) defines belief systems as “the explicit set of organizational definitions that senior managers communicate formally and reinforce systematically to provide basic values, purpose, and direction for the organization” (p.34). 7 Kraus et al. (2017) document how ideological controls involving repeated talks have an important impact on how organizational members perceive implementation of formal controls. 11 I argue that CEO activism is an informal communication of corporate values and beliefs that signals value alignments with existing and prospective employees. As articulated in Simons (1994), “a powerful way to signal commitment is through periodic, personal involvement in key aspects of [levers of control].” CEO activism often coincides with sociopolitical polarizing issues, enabling CEOs to take stances in a nuanced way; CEOs engaging in activism tend to incorporate pledges or proscriptive statements, pronouncing the degree to which they take the issues seriously. Lastly, CEO activism garners significant traction in the traditional or social media, ensuring the salience of its messages. Costs and Benefits of CEO Activism CEO activism is a form of nonfinancial, qualitative disclosure that signals the manager’s commitments to the values and beliefs the firm champions (Hambrick and Wowak 2021; Chen et al. 2022). However, one might question its credibility as a signal because CEO activism is almost always symbolic. The content of activism is often unrelated to the firm’s core businesses and is devoid of measurable actions. While messages delivered through CEO activism are rarely verified ex post or assured by a third party, I expect the presence of the less visible costs (e.g., damages to personal reputation, backlash from discontent stakeholders, potential political sanctions) to discipline the CEOs. Economic theories suggest underlying features of CEO activism make CEO activism generally credible. The cheap talk model in Farrell and Rabin (1996) maintains that unverifiable disclosures are relevant when agents’ incentives are at least partially aligned. Relatedly, reputational costs associated with CEO activism also bolster credibility of CEO activism. Stocken (2000) argues that in the setting of multi-period games, the manager has incentives to truthfully reveal private information and build a reputation of reporting truthfully. Kartik (2009) shows that 12 when lying is costly, the informed party will report with inflated language but not misreport. Incentives partially aligned between the CEO and employees (e.g., personal wealth derived from the employment) and repeated interactions suggest that CEO activism is a costly signal of the firm’s values and beliefs. Notwithstanding the costs discussed above, rational CEOs are likely to engage in activism when the expected benefits offset the expected costs. Prior theoretical work in voluntary disclosure suggests that managers have incentives to voluntarily disclose private information (i.e., stances on socially sensitive issues). Gigler (1994) constructs an analytical model to show that disclosure costs can increase disclosures by supplying credibility for unaudited, voluntary disclosures. Ferreira and Rezende (2007) argue that managers have incentives to voluntarily disclose private information about corporate strategy to stakeholders. These authors show that with the information revealed, stakeholders can make investments specific to strategic directions consistent with the manager’s disclosure. Hypotheses I develop three hypotheses to examine whether CEO activism functions as a value-based, informal control. First, I hypothesize that CEO activism is associated with a decline in the aggregate employee sentiment. Given its salience and its contentious nature, CEO activism will galvanize employees into a strong sense of support or distaste depending on whether individual employees agree with the content of CEO activism. Individuals generally dislike others or entities whose beliefs are incongruent with their own (Brewer 1999; Golman et al. 2016).8 Adding nuance to these 8 Brewer (1999) discusses individuals’ tendency to dislike outgroup members (“outgroup hate”) as described in prior studies. Golman et al. (2016) note that individuals have distaste for belief dissonance. These ideas are broader than the idea of homophily, which does not necessarily assume individuals’ discontent associated with outgroup members or differing beliefs. 13 theories, recent studies in political science document that individuals hold animosity toward others who have opposing political views, a phenomenon known as affective polarization (Iyengar and Westwood 2015; Iyengar et al. 2019; Orr and Huber 2020). CEOs who engage in activism, by disclosing their sociopolitical views, expose themselves to such animosity from employees. Prior studies also show that in the corporate setting where employees may agree or disagree with beliefs in general, aggregate employee sentiment is negative due to demotivating effects of sociopolitical statements (Burbano 2021) or negative perceptions toward others with an opposing stance (Dimant 2023). I formally state my first hypothesis in an alternative form, predicting negative effects of CEO activism on aggregate employee sentiment. H1: CEO activism negatively affects employee sentiment. The second hypothesis pertains to employee turnover. Theories in economics contend that (mis)alignment of identity is an important source of (dis)utility for individuals (Akerlof and Kranton 2000, 2005; Benabou and Tirole 2011). In the context of employment decisions, these studies show that employees self-sort based on how they identify with employers (Akerlof and Kranton 2005; Van den Steen 2005, 2010; Henderson and Van den Steen 2015). The (mis)alignment of identity and values is a strong driver of employment decisions in labor markets. Prior studies have amassed empirical evidence that political or partisan-based discrimination predicts employment, job search, and retention decisions of firms and individuals (Gift and Gift 2015; McConnell et al. 2018; Bermiss and McDonald 2018; Babenko et al. 2020; Roth et al. 2021; Colonnelli, Neto, and Teso 2022). These studies suggest that to the extent that it signals values and beliefs advocated in the firm, CEO activism will appeal to incumbent employees whose beliefs align with the CEO’s stance and alienate employees who identify against the stance. Following this logic, I expect CEO activism 14 to subsequently intensify the sensitivity of employee turnover to employee sentiment. I formally state the hypothesis as follows. H2: CEO activism amplifies the sensitivity of employee turnover to employee sentiment. For CEO activism to be a valid component of the control systems, the churning of employees should eventuate in observable firm-level benefits. My last hypothesis is that CEO activism is positively associated with the firm’s operating performance. The performance implications of CEO activism can manifest via different channels that are not mutually exclusive. On the one hand, a positive association between CEO activism and operating performance arises from greater employee productivity or efforts on average after the churning. Prior studies note that individuals who identify themselves with their employers are intrinsically more motivated and will exert greater effort (Van den Steen 2005; Besley and Ghatak 2005; Prendergast 2008). In addition, sorting employees via voluntary separations may create a relatively homogeneous culture within a firm, which prior studies find increases firm performance because of efficient coordination among employees (Van den Steen 2010; Corritore, Goldberg, and Srivastava 2020). Consistent with these two channels, I predict that CEO activism is positively associated with the subsequent operating performance of the firms. H3: CEO activism positively affects the firm’s operating performance. These hypotheses collectively allow me to probe the research question of whether CEO activism functions as an informal value-based control by facilitating employee sorting. 15 Data III. RESEARCH DESIGN The data used in this study are from three primary sources. I begin with 2,494 firms whose CEOs are listed in Execucomp from 2011 to 2022.9 I match these firms with the data universe of Revelio Labs, a commercial provider of large-scale labor market analytics collected from multiple third- party sources, including Glassdoor, LinkedIn, and corporate websites. I require firms to have an identical CIK number, identical stock exchange and ticker, or firm names with fuzzy-match score above 95 to be considered a successful match.10 Data on CEO Activism I hand-collect data on CEO activism using Factiva and X (formerly Twitter). Consistent with prior studies (Larcker et al. 2018; Bhagwat et al. 2020; Mkrtchyan, Sandvik, and Zhu 2023), I capture occurrences of CEO activism based on socio-politically contentious key words appearing in titles of news articles and posts on personal X accounts of CEOs in the sample. Operationally, I include news articles in my sample of CEO activism if one or more of the keywords appears in the article headline along with the firm name and its CEO’s last name. Similarly, I collect posts on CEOs’ personal X accounts that contain one or more of the keywords. The total of 665 cases of CEO activism by 136 firms’ CEOs between 2011 and 2021 remain in the final sample.11 I list notable examples of CEO activism in Appendix B. 9 I select Execucomp as the initial sample universe for three reasons. First, the Execucomp universe represents S&P 1500 firms and is sufficiently large to entice attention from the traditional and social media users. Second, because this study places an extensive emphasis on CEO behavior, using Execucomp ensures consistency in identifying CEO names and duration of their tours. Third, using firms in Execucomp results in reasonably good matches with Revelio Labs, whose coverage is sparse for smaller firms. 10 I manually verify the stock tickers to ensure that the matches are not capturing legacy tickers. 11 The sample period for CEO activism ends in 2021 to analyze effects of CEO activism on subsequent firm-level outcomes. 16 An important design choice in the collection of data on CEO activism is whether to screen the cases of CEO activism based on relations with the firm’s core business. While an ideal setting of CEO activism would contain only cases of CEO activism that are orthogonal to the firms’ businesses, scrutinizing ties between the CEO’s statement and the firms’ businesses involves significant subjectivity. Following Hambrick and Wowak (2021) and Mkrtchyan, Sandvik, and Zhu (2023), I incorporate cases of CEO activism irrespective of whether they are related to the firm’s businesses. Data on Employee Sentiment and Employee Turnover Data on employee sentiment comes from Revelio Labs, which compiles the raw data from Glassdoor.com. 12 Prior empirical studies commonly use employee reviews collected from an online platform as a reasonable proxy for employees’ aggregate perceptions of firms (Teoh 2018; Hales, Moon, and Swenson 2018; Huang, Li, and Markov 2020; Lee et al. 2021; Welch and Yoon 2022). Revelio Labs also compiles individuals’ LinkedIn profiles, enabling researchers to examine self-reported attributes (timeline, job title, firm, and location) of their jobs and some biographical data (education) at the individual episode level. Using self-reported profiles to infer non-executive employees’ employment changes has increasingly become common in recent accounting studies (Lee, Naiker, and Stewart 2022; Lang, Pinto, and Sul 2023). These data are initially at individual review (Glassdoor.com) or individual episode (LinkedIn) level, and I collapse the granular data to firm-month level for the firms retained from the previous stage. I then merge these firm-month observations with data on CEO activism and attach control variables constructed from Compustat Quarterly and CRSP. After I remove observations matched 12 Glassdoor.com is an anonymized online platform that enables employees to rate their firms on business outlook, career opportunities, culture and values, work/life balance, and other factors, and to post reviews about what they see as pros and cons of the firms. 17 with less than 50 employees and those with missing dependent or control variables, the final sample contains 59,163 firm-month observations across 1,220 unique firms. Using the same sample universe and a nearly identical sample construction process, I also construct a firm-year panel to specifically test effects of CEO activism on the firm’s subsequent operating performance (H3). 13 I shift from firm-month to firm-year observations for this test primarily because i) outsiders can observe only limited dimensions of firms’ monthly performance (e.g., monthly revenue at select retail stores), and ii) a firm’s monthly performance is affected heavily by noise arising from operational seasonality or external events. The final sample for the firm-year panel consists of 13,391 firm-year observations. Table 1 summarizes the sample construction processes. Dependent Variables In Section 2, I hypothesize that CEO activism i) negatively affects employee sentiment in the short run, ii) facilitates sorting employees, and iii) is positively associated with the firm’s subsequent operating performance in the longer run. Following prior studies that use Glassdoor.com reviews (Hales et al. 2018; Li et al. 2020), I employ the monthly mean of numerical ratings that employees leave as a part of the reviews to infer firms’ aggregate employee sentiment before and after occurrences of CEO activism. Specifically, I use ratings for the senior management (Mean_rating_management) and culture and values (Mean_rating_culture) to capture dimensions 13 The employee-level and ratings data from Revelio Labs end in September 2022. Merging this data with Compustat’s annual data results in missing observations for firms whose fiscal year ends after September. To avoid sample selection bias, the sample period for the firm-year panel concludes in 2021, ensuring that such firms are excluded. 18 of employee sentiment that relate closely to a firm’s values and beliefs, and overall ratings (Mean_rating_overall) to capture employees’ overall sentiment toward the firm.14,15 I measure the effects of CEO activism on employee sorting by examining employee turnover counts and rates at the firm-month level. Because Revelio Labs provides initial data at the individual episode level (usually corresponding to a single position held by a LinkedIn user) and many LinkedIn users disclose when they start and leave a position, I can obtain the number of employees at a firm on a given date.16 I calculate monthly employee turnover rate by dividing the number of U.S.-based employees that separate from a firm in a given month by the number of all U.S.-based employees who work for the firm averaged over the past 12 months and then multiplying by 100. 17 To proxy for employee sentiment, I employ i) moving averages of Mean_rating_overall over rolling six-month windows, standardized within Fama-French 48 industry-year, and ii) decile rankings of the proportion of employees who gave their firms very high (i.e., 5 out of 5) or very low (i.e., 1 out of 5) scores on the overall ratings. The use of separate measures for the very high and very low ratings helps distinguish how CEO activism differentially influences the relations between employee turnover and their positive and negative sentiment. For the test of H3, I utilize the year-ahead natural log of sales per employee (Log(sales_per_emp)) and return on assets (ROA) as proxies for the firms’ operating performance. 14 I exclude reviews from users who do not claim to be current employees. 15 As an additional analysis, I examine effects of CEO activism on ratings on Glassdoor.com unrelated to beliefs and values in Section 5. 16 Agrawal, Hacamo, and Hu (2021) provide survey evidence that 76% of LinkedIn users update their profiles within a month of starting a new position. I require episodes to have at least one of the dates to be included in the analyses. 17 This approach is similar to that used in Li et al. (2022), who also calculate employee outflow rates using Revelio Labs data, and the estimated turnover rates are comparable. These authors further validate the turnover rate by exhibiting positive correlations between Revelio Labs-based turnover rate and turnover data reported by Bureau of Labor Statistics (BLS). 19 Both measures are widely used as firm-level measures of operating performance, and the use of these proxies enables me to examine both productivity and profitability, important yet sufficiently distinct aspects of firms’ operating performance. Empirical Design I incorporate the following OLS specification with two-way fixed effects to examine H1: Yf,m,y = β0 + β1Activism [1 month]f,m,y + ΣControls + firm FE + year FE + month FE + ϵf,m,y (1) where 𝑌𝑓,𝑚,𝑦 denotes the monthly mean of one of the employee sentiment measures for firm 𝑓 in month 𝑚 of fiscal year 𝑦 ; 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [1 𝑚𝑜𝑛𝑡ℎ]𝑓,𝑚,𝑦 is a binary indicator set to one for firm- month observations that are within a month after firm 𝑓’s CEO engaged in CEO activism and zero otherwise. H1 predicts 𝛽1 < 0. For the test of H2, I employ two modified versions of equation (1) to analyze the sensitivity of employee turnover to employee sentiment: Yf,m,y = γ0 + γ1Activism [6 months]f,m,y + γ2Sentimentf,m,y + γ3Activism [6 months]f,m,y × Sentimentf,m,y + ΣControls + firm FE + year FE + month FE + vf,m,y (2) Yf,m,y = θ0 + θ1Activism [6 months]f,m,y + θ2Pos_sentf,m,y + θ3Neg_sentf,m,y + θ4Activism [6 months]f,m,y × Pos_sentf,m,y + θ5Activism [6 months]f,m,y × Neg_sentf,m,y + ΣControls + firm FE + year FE + month FE + wf,m,y (3) where 𝑌𝑓,𝑚,𝑦 represents the monthly measure of employee turnover for firm 𝑓 in month 𝑚 of fiscal year 𝑦; 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [6 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 is a binary indicator set to one for firm-month observations that are within six months after firm 𝑓’s CEO engaged in CEO activism and zero otherwise. H2 20 predicts 𝛾3 < 0 in equation (2) and 𝜃4 < 0 and 𝜃5 > 0 in equation (3).18 I include a set of control variables broadly consistent with the prior studies examining employee sentiment (Hales et al. 2018; Lee et al. 2021).19 Table A1 illustrates the identification strategy. The primary purpose of the above specification is to compare the dependent variables post- CEO activism against the pre-CEO activism period within the same firm and the control firms. I use the treatment window of a month (six months) for the test of H1 (H2) for two important reasons. First, the sample period used in this study spans 12 years, and CEOs engage in CEO activism at different times. This feature of the setting requires me to pinpoint treatment windows rather than simply bisect the entire sample period to pre- and post-CEO activism periods. Second, varying the treatment windows across the tests helps juxtapose a short-term effect of CEO activism on employee sentiment and its impact on employee sorting in the longer term. Figure 2 visualizes the timeline of these analyses. The test of H3 utilizes a different unit of analysis and uses the following OLS specification: Performancef,y+1 = δ0 + δ1Activism [yearly]f,,y + ΣControls + firm FE + year FE + uf,,y (4) where 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑓,𝑦+1 refers to one-year-ahead firm 𝑓 ’s operating performance, and 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [𝑦𝑒𝑎𝑟𝑙𝑦]𝑓,𝑦 is a binary dummy set to one if firm 𝑓’s CEO engaged in activism in fiscal year 𝑦 and zero otherwise. Control variables are the same as those used in equations (1) to (3). H3 predicts 𝛿1 > 0. 18 An underlying assumption in H2 is that employee sentiment is negatively associated with employee turnover. That is, I expect 𝛾2 < 0, 𝜃2 < 0, and 𝜃3 > 0. 19 Control variables are the natural log of total assets (Size), market-to-book ratio (MB), return on assets (ROA), financial leverage (Leverage), proportion of firms’ cash holdings (Cash), stock return (BHR), and return volatility (RetVol). I winsorize all continuous variables at the 1st and 99th percentiles. Appendix A defines all variables. 21 Overview of CEO Activism IV. RESULTS I provide an overview of occurrences of CEO activism in Table 2 by year and month (Panel A), industry (Panel B), topic (Panel C), and frequency (Panel D).20 In Panel A, I decompose the list of CEO activism by year and month to demonstrate the rapid increase in its occurrences particularly since 2017. The highest concentration of CEO activism in the sample occurs in June 2020 (32 cases), which concurs with the murder of George Floyd (May 25, 2020). Its temporal dispersion indicates that CEO activism closely postdates major sociopolitical events (e.g., inauguration of Donald Trump, attack on U.S. Capitol). Panel B of Table 2 exhibits occurrences of CEO activism by Fama-French 12-industry classifications and compares it with the composition of each industry in the Execucomp universe. The results show that the highest proportion of CEO activism occurs in the business equipment industry (30.38% of the total CEO activism cases). Panel C shows that diversity and politics tend to be the most common topic of CEO activism.21 Panel D of Table 2 reports the number of firms by the frequency of CEO activism throughout the sample period. The distribution shows that 11.15% of the sample firms (136 out of 1,220) are involved in CEO activism at least once during the sample period. Table 3 reports descriptive statistics for the firm-month panel used to test H1 and H2. Panel A of Table 3 reports summary statistics for the treatment variables, dependent variables, and control variables. Distributions of the dependent variables generally accord well with prior studies that use Glassdoor ratings (Hales et al. 2018) and LinkedIn-based Revelio Labs data (Li et al. 20 Appendix C lists keywords used to identify CEO activism by topic. 21 The topics are not mutually exclusive, and therefore the percentages add up to more than 100%. However, multi-topic activism is not common in my sample: more than 90% of the cases involve only one topic. 22 2022). Panel B of Table 3 displays tests of mean differences in key variables between firms whose CEOs never engaged in CEO activism during the sample period (“Activism – never”) and firms whose CEOs did so at least once (“Activism – at least once”). The univariate tests reveal that CEO activism occurs at significantly larger firms. Panel C of Table 3 reports pairwise Pearson correlations among the variables. The correlations show that firms involved in CEO activism receive more favorable ratings from employees, lower turnover rate of employees, are larger, and demonstrate less volatile returns in the stock market. Given the insignificant correlation between CEO activism and stock return (BHR), firm performance in the stock market is unlikely to be among key drivers of CEO activism. I discuss how CEO activism is tied to stock market responses in the following subsection. Stock Market Reactions to CEO Activism I begin my analyses by examining stock market reactions to CEO activism. Doing so validates CEO activism in two critical ways. First, demonstrating value-relevance of CEO activism elucidates that that CEO activism has information content and therefore constitutes a meaningful signal to stakeholders. Second, examining abnormal returns over a short-term window surrounding CEO activism helps isolate CEO activism from confounding events that coincide with occurrences of CEO activism. Prior studies document mixed stock market reactions to CEO activism and firms’ sociopolitical engagements (Bhagwat et al. 2020; Bedendo and Siming 2021; Chen et al. 2022; Mkrtchyan, Sandvik, and Zhu 2023; Brownen-Trinh and Orujov 2023). The contentious nature of CEO activism implies that firms whose CEOs engage in CEO activism are subject to risks 23 associated with support and/or backlash from stakeholders.22 To the extent that investors view such risks as value-relevant, it is reasonable to expect negative stock market reactions to CEO activism, at least in the short run. I measure stock market reactions to CEO activism by calculating daily abnormal returns based on time-series estimations regressing a firm’s daily raw returns on market returns (Campbell, Lo, and MacKinlay 1998; Acemoglu et al. 2016). Specifically, I decompose the daily raw return for firm 𝑖 on trading day 𝑡 (𝑅𝑖𝑡) into predicted return (𝑅𝑖𝑡 ∗ ) and abnormal return (𝐴𝑅𝑖𝑡) and estimate ∗ as a linear function of the market return: 𝑅𝑖𝑡 Ri,t = 𝑅𝑖,𝑡 ∗ + ARi,t = αi + βi Rm,t + ϵi,t (5) where 𝑅𝑚,𝑡 denotes the raw return on S&P 500 index on trading day 𝑡. I estimate the parameters 𝛼𝑖 and 𝛽𝑖 on pre-event rolling windows spanning 250 days ending 30 days prior to CEO activism.23 The residual estimated from the above specification (𝜖̂𝑖,𝑡) represents firm 𝑖’s abnormal return on day 𝑡, which shows how a firm’s daily raw returns compare with its own performance in the market over the 250-day estimation period. The baseline findings reveal that occurrences of CEO activism are associated with negative short-term reactions in the stock market. Table 4 compares the mean of daily abnormal returns of firms whose CEOs engage in activism (“activism firms”) and other firms in the sample (“control 22 For example, anecdotal evidence suggests that consumer boycotts are a common consequence of CEO activism. Empirical studies examining sales implications of CEO activism report its negative impacts on sales (Hou and Poliquin 2023; Jin et al. 2024). Conversely, CEO activism may garner support from institutional investors emphasizing firms’ social initiatives. 23 The market model used here has two advantages. First, using within-firm time-series estimations of abnormal returns accounts for confounding factors in the stock market. Second, this specification does not rely on backward-looking accounting information. The pre-event window of 250 trading days follows the prior studies and roughly corresponds to a year of trading (Li and Lie 2006; Acemoglu et al. 2016). 24 firms”) for 11 days surrounding CEO activism denoted as event day 0.24 Notably, activism firms experience significantly lower abnormal returns on trading days immediately following occurrences of CEO activism (i.e., event day 1) (t = 2.02) and none of the other days prior to CEO activism. The negative market reactions to CEO activism appear short-lived, as the reactions subsequently dissipate and reverse on event day 5. Figure 1 depicts the daily abnormal returns of activism and control firms over an 11-day window surrounding CEO activism.25 The tests of stock market reactions indicate that the stock market perceives CEO activism negatively in the short run, substantiating the idea that CEO activism is a costly signal.26 Test of H1 I examine H1 by estimating equation (1) using Mean_rating_management, Mean_rating_culture, and Mean_rating_overall as dependent variables. I employ the treatment window of one month, meaning that the estimated coefficient on 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [1 𝑚𝑜𝑛𝑡ℎ]𝑓,𝑚,𝑦 captures the difference in dependent variables i) between activism firms and control firms in the month following CEO activism and ii) within activism firms across the post-activism month and other months. Table 5 reports the result of this estimation. Consistent with H1, the coefficient on 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [1 𝑚𝑜𝑛𝑡ℎ] in Panel A is significantly negative. Column (1) shows that activism firms 24 For cases of CEO activism that occur on nontrading days, I omit event day 0 and code the first subsequent trading day as event day 1. 25 Unreported tests on firms’ cumulative abnormal returns (CARs) also show that CARs to activism firm remain lower relative to control firms for four days and display no statistically significant difference beginning on event day 5. 26 The negative stock market reactions reported here seemingly contrast with those of Mkrtchyan, Sandvik, and Zhu (2023). This discrepancy likely arises from differences in sample selection procedures and return measures. Replicating their paper using their sample period from 2011 to 2019 and their measure of market-adjusted stock market returns, I observe a pattern similar to those reported in Table 2 of their paper (positive and significant returns on the event days up to event day 2). 25 see Mean_rating_management decline by 0.093 points (t = -2.19) after controlling for time-variant attributes. Such levels of decline in employee morale are economically significant: given that the mean of Mean_rating_management is 2.867, the drop is equivalent to 3.24% of the mean rating for senior management across the sample. Column (3) [column (5)] exhibits similar effects of CEO activism on mean ratings for culture and values [overall ratings] (t = -1.77 [t = -1.64]). These findings suggest that employees express meaningful discontent about the firm’s management and values, and, to a lesser degree, toward the firm itself, following CEO activism. Columns (2), (4), and (6) show that these findings are robust to inclusion of time-variant control variables at the firm-quarter level.27 Importantly, the specification design used in this test provides some assurance regarding concerns of endogeneity. First, these results hold in the presence of firm, year, and month fixed effects, effectively controlling for time-invariant unobservable factors, year-by-year macroeconomic factors, and seasonality in firms’ operations, respectively. Second, using the post-activism one-month window helps rule out alternative interpretations that a long-term time trend drives the employee sentiment. I report results of month- by-month dynamic analyses in Panel A, Table A2. A logical follow-up question is whether CEO activism discussing different topics has varying effects on employee sentiment. In Panel B, Table 5, I further decompose 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [1 𝑚𝑜𝑛𝑡ℎ] into six binary dummy variables each representing a topical group, as listed in Appendix C. Interestingly, CEO activism on politics has negative effects on subsequent employee ratings, while activism on other topics appear to have mixed or insignificant effects. For a firm with mean ratings 27 The control variables used in this estimation are measured at the firm-quarter levels and rarely change during the treatment windows for CEO activism. Including control variables in columns (2), (4), and (6) therefore makes only trivial changes to the coefficient estimates. 26 across all three rating dimensions, CEO activism on politics is associated with declines of 3.83% to 5.51% in the ratings. Test of H2 I now utilize equations (2) and (3) using Log(turnover_count) and Turnover_rate as dependent variables to test H2. Here I examine effects of CEO activism on the sensitivity of employee turnover to employee sentiment, using the treatment window of six months following occurrences of CEO activism.28 Table 6 reports the results of this test. In panel A, I proxy for employee sentiment by incorporating a singular measure of employee sentiment (Emp_sent) (columns (1) and (2)) and a dual-measure of employee sentiment separately measured for positive (Emp_sent_pros_overall) and negative (Emp_sent_cons_overall) sentiment. In column (1), the estimated coefficient on 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [6 𝑚𝑜𝑛𝑡ℎ𝑠] is 0.040 (t = 2.26), meaning that for a firm with average employee sentiment (i.e., 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡 = 0 ), CEO activism is associated with an increase in employee turnover. The coefficient on 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡 is -0.018 (t = -6.46), indicating that without CEO activism, a unit increase in employee sentiment is associated with a 1.8-percentage-point decrease in employee turnover. The coefficient on the interaction term 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [6 𝑚𝑜𝑛𝑡ℎ𝑠] × 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡, which is of interest, is −0.048 . This result shows that with CEO activism, a unit increase in 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡 is now associated with a 6.6-percentage-point decrease in employee turnover. Column (2) employs separate proxies for employees’ positive and negative sentiments. The coefficient estimate on the interaction term 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [6 𝑚𝑜𝑛𝑡ℎ𝑠] × 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑝𝑟𝑜𝑠_𝑜𝑣𝑒𝑟𝑎𝑙𝑙 is -0.009 (t = -2.18), amplifying the main effect of 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑝𝑟𝑜𝑠_𝑜𝑣𝑒𝑟𝑎𝑙𝑙 on employee turnover (-0.004, t = 28 Given that employment decisions are individuals’ major economic decisions that can take considerable time, imposing a short treatment window to test consequences of CEO activism on employee turnover would be unreasonable. 27 -3.95). Results in columns (3) and (4) are qualitatively similar: CEO activism intensifies the sensitivity of turnover rate to employee sentiment by more than three-fold (column 3) and significantly increases turnover rate in firms where more employees hold negative sentiment toward their firms (column 4). These results are generally consistent with H2, which predicts that CEO activism amplifies the sensitivity of employee turnover to employee sentiment. Importantly, these results are of great economic significance: CEO activism amplifies the sensitivity of employee turnover to employee sentiment by 125% to 357%. The strong magnitudes indicate that CEO activism significantly facilitates sorting of employees by amplifying (attenuating) employee turnover for employees with negative (positive) sentiment toward their firms. Panel B, Table 6 analyzes the sensitivity of employee turnover to employee sentiment regarding the top management (columns (1) and (2)), firm culture (columns (3) and (4)), and politics (columns (5) and (6)). Across all columns, the results show that absent CEO activism, employees’ negative sentiment in any of the three dimensions is strongly associated with employee turnover. Interestingly, I find that CEO activism significantly intensifies the sensitivity of employee turnover to employees’ negative sentiment in firm culture (t = 2.62 to 2.79) and politics (t = 2.22 to 2.63), but not as strongly in the top management (t = 1.52 to 1.72). The findings in Table 6 collectively suggest that while CEO activism generally strengthens the relationships between employee sentiment and employee turnover, the most substantial sorting of employees manifests when CEO activism exacerbates employees’ negative sentiment towards the firm’s culture or political stances. Test of H3 The tests in the previous subsections examine how employees respond to CEO activism at the aggregate level. In this subsection, I examine the firm-level performance implications of CEO 28 activism, attempting to present a more complete picture of the rationale that CEO activism serves as an informal control. Specifically, I employ equation (4) at the firm-year level to test how CEO activism affects firms’ productivity (Log(Sales_per_emp)) and profitability (ROA) in the subsequent year. Table 7 reports descriptive statistics of the firm-year panel used in this test and the results of estimating equation (4). Panel A (Panel B) reports summary statistics (Pearson correlation matrix) for the variables used in this analysis, and Panel C presents the results of the OLS specification testing H3. Here the variable of interest is 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚𝑓𝑦, the binary indicator set to one if the firm’s CEO engaged in CEO activism during the fiscal year and zero otherwise. I find results consistent with H3. Column (1) [column (2)] shows that occurrence of CEO activism is positively associated with one-year-ahead Log(Sales_per_emp) [ROA] (t = 2.00 [t = 2.02]). The economic significances of these coefficient estimates are not trivial. The coefficient on column (2) corresponds to an increase of 11.36% for firms with the mean ROA. These findings suggest that CEO activism is positively associated with the firm’s subsequent productivity and profitability. The empirical results described thus far report that CEO activism results in negative market reactions (Table 4) and employee sentiment (Table 5) in the short run and increased employee turnover (Table 6) in the longer run. CEO activism is also positively associated with the firm’s year-ahead operating performance (Panel C, Table 7). I argue that when collated, these findings jointly indicate that CEO activism functions as an informal control by polarizing employees based on value alignments. 29 V. ADDITIONAL ANALYSES A key criticism of the argument that CEO activism functions as an informal control is that I do not observe direct links among employee sentiment, employee churning, and the firm’s operating performance. The argument is then subject to an alternative explanation that different unobservable factors lead CEOs to engage in activism, degrade employee sentiment, facilitate employee sorting, and improve the firm’s operating performance in a sequential manner. Therefore in this section, I present results of additional analyses to supplement the main findings discussed in Section 4 and rule out alternative explanations. Employees’ Ratings Unrelated to Values and Beliefs An alternative explanation that would erode the effects of CEO activism on employee sentiment is that employees react negatively to CEO activism because of reasons unrelated to values and beliefs. For example, employees at an activism firm may ex ante fear economic consequences of CEO activism (e.g., declining sales due to consumer boycotts, fear of repeated negative stock market reactions). To the extent that such fear dominates their overall perception toward the firm or spills over to their perception of the corporate culture and values, the effects of CEO activism on employee sentiment reported in Table 5 may be overestimated. I examine the effects of CEO activism on other components of employee sentiment to address the above suspicion. Glassdoor.com collects employees’ ratings of their companies on multiple factors including career opportunities, work-life balance, and the firms' business outlooks. If CEO activism affects employee sentiment through the communicated values and beliefs but not through adjustments in other aspects of employees’ perceptions toward firms, I should observe no significant associations between CEO activism and other ratings on Glassdoor.com. Table 8 reports the results of the above estimations. I find no significant evidence that CEO activism affects 30 employee sentiment on career opportunities (t = -0.99), work-life balance (t = -1.04), or the firms' business outlooks (t = -1.00). These results, combined with the main findings of Panel A, Table 5, bolster the argument that CEO activism communicates values and beliefs and thereby affects employee sentiment. Longer-term Effects of CEO Activism on Employee Sentiment A question unanswered in Section 4 is whether activism firms suffer from poor employee sentiment for an extended period. In such a case, the improvements in operating performance following CEO activism would be difficult to reconcile with the sustained decline in employee sentiment. To address this point, I regress the measures of employee sentiment used in Panel A, Table 5, varying the window of post-activism period. In untabulated results, I find no significant evidence that CEO activism negatively affects any measure of employee sentiment over the three-, six-, or 12-month windows following CEO activism. The lack of significant long-term negative effects of CEO activism on employee sentiment suggests that the above concern is unlikely to hold. Recruiting New Employees I show in Section 4 that CEO activism is associated with an increased sorting of employees and the firm’s improved operating performance. However, these results are limited in describing mechanisms through which the churning of employees is associated with activism firms’ operating performance. Without the mechanism, the findings of this paper are not free from an alternative explanation that activism firms subsequently shrink and yet become more profitable by offloading unproductive employees. To address the above criticism, I estimate the effect of CEO activism on the firm’s subsequent number of employees and find no significant association (untabulated). I next analyze how successful activism firms are in recruiting new employees following CEO activism. I estimate 31 equations (2) and (3) using measures of employee recruiting as dependent variables and report the results in Table 9. In panel A of Table 9, I show the effect of CEO activism on the sensitivity of employee recruiting to employee sentiment, analogous to Table 6. I find that the singular measure of employee sentiment (𝐸𝑚𝑝_𝑠𝑒𝑛𝑡 ) does not significantly affect the sensitivity of recruiting to employee sentiment (columns (1) and (3)). However, columns (2) and (4) indicate that CEO activism is associated with an increase in recruiting (t = 2.01 to 2.10) when the employee sentiment is neither positive nor negative. As a caveat, CEO activism also attenuates the sensitivity of recruiting to employees’ positive and negative sentiment. These results suggest that CEO activism may be effective in facilitating recruiting in some cases. A follow-up question is whether CEO activism can be useful in attracting a specific subset of labor market participants. I follow Agrawal et al. (2021) to further test cross-sections of the recruited employees based on their education. In panel B, Table 9, I report effects of CEO activism on the firm’s recruiting separately across the new employees’ level of education. Interestingly, I find some evidence in columns (4) [column (6)] that CEO activism significantly facilitates recruiting of new employees with a bachelor’s [graduate] degree, in the absence of employees’ strong positive or negative sentiment (t = 2.35) [t = 2.89]. However, these results suggest different implications for the sensitivity of recruits to employee sentiment: CEO activism offsets (exacerbates) the relationships between employees’ positive (negative) sentiment and recruits. With the caveat, these results suggest that CEO activism is somewhat effective in recruiting highly educated employees, who in turn may be responsible for the improved operating performance at activism firms. Also, one should exercise caution in interpreting these results, as these results are unable to pinpoint employee sorting at individual position level. 32 VI. CONCLUSION I posit that CEO activism functions as an informal control mechanism, facilitating employee turnover based on alignment with corporate values. Distinct from CSR activities and firms’ political contributions, CEO activism is an emerging phenomenon that is highly visible and is often polarizing in nature. I draw from the “levers of control” framework (Simons 1995) to argue that CEOs use activism to informally communicate values and beliefs to employees, signaling commitments to the values. To operationalize the argument, I hypothesize and empirically show that CEO activism negatively affects employee sentiment in the short run and amplifies the sensitivity of employee turnover to employee sentiment. In the longer run, I also find that CEO activism is positively associated with the firm’s year-ahead operating performance. I corroborate these findings by showing that CEO activism facilitates the firm’s recruiting absent strongly polarized employee sentiment, as labor market participants holding more advanced degrees are willing to join the activism firms following occurrences of CEO activism. While I do not observe direct causal ties among these outcomes, the sequence of analyses collectively suggests that CEO activism serves as a component of a firm’s belief system that entails both significant costs in the short run (negative market reactions, declining employee sentiment, higher turnover of employees with negative sentiment) and benefits in the longer run (retention of employees with positive sentiment, hiring benefits, and improved operating performance). This paper sheds light on CEO activism as a control mechanism that managers can employ to seek value alignment with employees. Importantly, it bridges the emerging literature on CEO activism with the management accounting literature on firms’ use of informal control components, especially those involving informal communications (Guiso et al. 2015b; Kachelmeier et al. 2016). 33 In doing so, this study responds to Van der Stede's (2011) call for management accounting research that examines how disclosures relate to firms’ internal operations. 34 REFERENCES Acemoglu, D., S. Johnson, A. Kermani, J. Kwak, and T. Mitton. 2016. The value of connections in turbulent times: Evidence from the United States. Journal of Financial Economics 121 (2): 368–391. Agrawal, A., I. Hacamo, and Z. Hu. 2021. Information dispersion across employees and stock returns. The Review of Financial Studies 34 (10): 4785–4831. Akerlof, G. A., and R. E. Kranton. 2000. Economics and identity. The Quarterly Journal of Economics 115 (3): 715–753. Akerlof, G. A., and R. E. Kranton. 2005. Identity and the Economics of Organizations. Journal of Economic Perspectives 19 (1): 9–32. Akinyele, K. O., V. Arnold, and S. G. Sutton. 2020. Wording and saliency matter: The impact of incentive system and organizational value statement on employees’ performance. Behavioral Research in Accounting 32 (1): 101–118. Babenko, I., V. Fedaseyeu, and S. Zhang. 2020. Do CEOs affect employees’ political choices? The Review of Financial Studies 33 (4): 1781–1817. Bartkus, B. R., and M. Glassman. 2008. Do firms practice what they preach? The relationship between mission statements and stakeholder management. Journal of Business Ethics 83: 207–216. Bedendo, M., and L. Siming. 2021. To advocate or not to advocate: Determinants and financial consequences of CEO activism. British Journal of Management 32 (4): 1062–1081. Bénabou, R., and J. Tirole. 2011. Identity, morals, and taboos: Beliefs as assets. The Quarterly Journal of Economics 126 (2): 805–855. Bermiss, Y. S., and R. McDonald. 2018. Ideological misfit? Political affiliation and employee departure in the private-equity industry. Academy of Management Journal 61 (6): 2182–2209. Besley, T., and M. Ghatak. 2005. Competition and incentives with motivated agents. American Economic Review 95 (3): 616–636. Bhagwat, Y., N. L. Warren, J. T. Beck, and G. F. Watson IV. 2020. Corporate sociopolitical activism and firm value. Journal of Marketing 84 (5): 1–21. Boxell, L., Gentzkow, M., and Shapiro, J. M. 2017. Greater Internet use is not associated with faster growth in political polarization among US demographic groups. Proceedings of the National Academy of Sciences, 114 (40), 10612–10617. Brewer, M. B. 1999. The psychology of prejudice: Ingroup love and outgroup hate? Journal of Social Issues 55 (3): 429–444. 35 Brownen-Trinh, R., and A. Orujov. 2023. Corporate socio-political activism and retail investors: Evidence from the Black Lives Matter campaign. Journal of Corporate Finance 80: 102417. Burbano, V. C. 2021. The demotivating effects of communicating a social-political stance: Field experimental evidence from an online labor market platform. Management Science 67 (2): 1004–1025. Cai, W. 2023. Formalizing the Informal: Adopting a Formal Culture-fit Measurement System in the Employee Selection Process. The Accounting Review. Campbell, J. Y., A. W. Lo, A. C. MacKinlay, and R. F. Whitelaw. 1998. The econometrics of financial markets. Macroeconomic Dynamics 2 (4): 559–562. Chatterji, A. K., and M. W. Toffel. 2018. The new CEO activists. HBR’S 10 MUST. Chatterji, A. K., and M. W. Toffel. 2019. Assessing the impact of CEO activism. Organization & Environment 32 (2): 159–185. Chen, A. J., P. Dechow, and S. T. Tan. 2022. Corporate Response to the Black Lives Matter Movement: Determinants of Speaking Out in Support of Social Causes. Available at SSRN 3921985. Christensen, D. M., H. Jin, S. A. Sridharan, and L. A. Wellman. 2022. Hedging on the Hill: Does Political Hedging Reduce Firm Risk? Management Science 68 (6): 4356–4379. Colonnelli, E., V. P. Neto, and E. Teso. 2022. Politics at work. National Bureau of Economic Research Working Paper. Corritore, M., A. Goldberg, and S. B. Srivastava. 2020. Duality in diversity: How intrapersonal and interpersonal cultural heterogeneity relate to firm performance. Administrative Science Quarterly 65 (2): 359–394. Dhaliwal, D. S., O. Z. Li, A. Tsang, and Y. G. Yang. 2011. Voluntary nonfinancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting. The Accounting Review 86 (1): 59–100. Dimant, E. 2023. Hate Trumps Love: The Impact of Political Polarization on Social Preferences. Management Science 70 (1): 1–31. Durney, M. T., J. Johnson, R. K. Sinha, and D. Young. 2022. CEO (in)activism and investor decisions. Available at SSRN 3604321. Edelman. 2022. Trust Barometer 2022. https://edl.mn/3rqOlx0. Accessed on January 8, 2024. Elliott, W. B., S. M. Grant, and F. D. Hodge. 2018. Negative news and investor trust: The role of $ Firm and × CEO Twitter use. Journal of Accounting Research 56 (5): 1483–1519. Farrell, J., and M. Rabin. 1996. Cheap talk. Journal of Economic Perspectives 10 (3): 103–118. 36 Ferreira, D., and M. Rezende. 2007. Corporate strategy and information disclosure. The RAND Journal of Economics 38 (1): 164–184. Gift, K., and T. Gift. 2015. Does politics influence hiring? Evidence from a randomized experiment. Political Behavior 37: 653–675. Gigler, F. 1994. Self-enforcing voluntary disclosures. Journal of Accounting Research 32 (2): 224– 240. Golman, R., G. Loewenstein, K. O. Moene, and L. Zarri. 2016. The preference for belief consonance. Journal of Economic Perspectives 30 (3): 165–188. Grabner, I., and F. Moers. 2013. Management control as a system or a package? Conceptual and empirical issues. Accounting, Organizations and Society 38 (6–7): 407–419. Graham, J. R., J. Grennan, C. R. Harvey, and S. Rajgopal. 2022. Corporate culture: Evidence from the field. Journal of Financial Economics 146 (2): 552–593. Guiso, L., P. Sapienza, and L. Zingales. 2015a. The value of corporate culture. Journal of Financial Economics 117 (1): 60–76. Guiso, L., P. Sapienza, and L. Zingales. 2015b. Corporate culture, societal culture, and institutions. American Economic Review 105 (5): 336–339. Gupta, A., and V. F. Misangyi. 2018. Follow the leader (or not): The influence of peer CEOs’ characteristics on interorganizational imitation. Strategic Management Journal 39 (5): 1437– 1472. Hales, J., J. R. Moon Jr, and L. A. Swenson. 2018. A new era of voluntary disclosure? Empirical evidence on how employee postings on social media relate to future corporate disclosures. Accounting, Organizations and Society 68: 88–108. Hambrick, D. C., and A. J. Wowak. 2021. CEO sociopolitical activism: A stakeholder alignment model. Academy of Management Review 46 (1): 33–59. Henderson, R., and E. Van den Steen. 2015. Why do firms have “purpose”? The firm’s role as a carrier of identity and reputation. American Economic Review 105 (5): 326–330. Homroy, S., and S. Gangopadhyay. 2023. Strategic CEO Activism in Polarized Markets. Journal of Financial and Quantitative Analysis: 1–74. Hou, Y., and C. W. Poliquin. 2023. The effects of CEO activism: Partisan consumer behavior and its duration. Strategic Management Journal 44 (3): 672–703. Huang, K., M. Li, and S. Markov. 2020. What do employees know? Evidence from a social media platform. The Accounting Review 95 (2): 199–226. 37 Iyengar, S., Y. Lelkes, M. Levendusky, N. Malhotra, and S. J. Westwood. 2019. The origins and consequences of affective polarization in the United States. Annual Review of Political Science 22: 129–146. Iyengar, S., and S. J. Westwood. 2015. Fear and loathing across party lines: New evidence on group polarization. American Journal of Political Science 59 (3): 690–707. Jin, H., K. Merkley, A. Sharma, and K. Ton. 2024. Customers’ response to firms’ disclosure of social stances: evidence from voting reform laws. Review of Accounting Studies. Kachelmeier, S. J., T. A. Thornock, and M. G. Williamson. 2016. Communicated values as informal controls: Promoting quality while undermining productivity? Contemporary Accounting Research 33 (4): 1411–1434. Kartik, N. 2009. Strategic communication with lying costs. The Review of Economic Studies 76 (4): 1359–1395. Kempf, E., Fos, V., and Tsoutsoura, M. 2023. The Political Polarization of Corporate America. Chicago Booth Research Paper, 22–14. Kraus, K., C. Kennergren, and A. von Unge. 2017. The interplay between ideological control and formal management control systems – A case study of a non-governmental organisation. Accounting, Organizations and Society 63: 42–59. Lang, M., J. Pinto, and E. Sul. 2023. MiFID II unbundling and sell-side analyst research. Journal of Accounting and Economics: 101617. Larcker, D. F., S. Miles, B. Tayan, and K. Wright-Violich. 2018. The double-edged sword of CEO activism. Rock Center for Corporate Governance at Stanford University Closer Look Series: Topics, Issues and Controversies in Corporate Governance No. CGRP-74, Stanford University Graduate School of Business Research Paper. Lee, G., V. Naiker, and C. R. Stewart. 2022. Audit office labor market proximity and audit quality. The Accounting Review 97 (2): 317–347. Lee, Y., S. Ng, T. Shevlin, and A. Venkat. 2021. The effects of tax avoidance news on employee perceptions of managers and firms: Evidence from Glassdoor. com ratings. The Accounting Review 96 (3): 343–372. Li, Q., B. Lourie, A. Nekrasov, and T. Shevlin. 2022. Employee turnover and firm performance: Large-sample archival evidence. Management Science 68 (8): 5667–5683. Li, W., and E. Lie. 2006. Dividend changes and catering incentives. Journal of Financial Economics 80 (2): 293–308. Malmi, T., and D. A. Brown. 2008. Management control systems as a package—Opportunities, challenges and research directions. Management Accounting Research 19 (4): 287–300. 38 McConnell, C., Y. Margalit, N. Malhotra, and M. Levendusky. 2018. The economic consequences of partisanship in a polarized era. American Journal of Political Science 62 (1): 5–18. Miller, G. S., and D. J. Skinner. 2015. The evolving disclosure landscape: How changes in technology, the media, and capital markets are affecting disclosure. Journal of Accounting Research 53 (2): 221–239. Mkrtchyan, A., Sandvik, J., and Xu, D. 2024. Employee Responses to CEO Activism. Available at SSRN 4506862. Mkrtchyan, A., J. Sandvik, and V. Z. Zhu. 2023. CEO Activism and Firm Value. Management Science. Orr, L. V, and G. A. Huber. 2020. The policy basis of measured partisan animosity in the United States. American Journal of Political Science 64 (3): 569–586. Pacelli, J., T. Shi, and Y. Zou. 2022. Communicating Corporate Culture in Labor Markets: Evidence from Job Postings. Available at SSRN 4235342. Prendergast, C. 2008. Intrinsic motivation and incentives. American Economic Review 98 (2): 201– 205. Roth, P. L., J. D. Arnold, H. J. Walker, L. Zhang, and C. H. Van Iddekinge. 2021. Organizational political affiliation and job seekers: If I don’t identify with your party, am I still attracted? Journal of Applied Psychology. Schultz, H. 2017. Message from Howard Schultz to Starbucks Partners: Living Our Values in Uncertain Times. https://stories.starbucks.com/stories/2017/living-our-values-in-uncertain- times/. Simons, R. 1994. How new top managers use control systems as levers of strategic renewal. Strategic Management Journal 15 (3): 169–189. Simons, R. 1995. Levers of control: How managers use innovative control systems to drive strategic renewal. Harvard Business Press. Stocken, P. C. 2000. Credibility of voluntary disclosure. The RAND Journal of Economics: 359– 374. Teoh, S. H. 2018. The promise and challenges of new datasets for accounting research. Accounting, Organizations and Society 68: 109–117. Tessier, S., and D. Otley. 2012. A conceptual development of Simons’ Levers of Control framework. Management Accounting Research 23 (3): 171–185. Van der Stede, W. A. 2011. Management accounting research in the wake of the crisis: Some reflections. European Accounting Review 20 (4): 605–623. 39 Van den Steen, E. 2005. Organizational beliefs and managerial vision. Journal of Law, Economics, and organization 21 (1): 256–283. Van den Steen, E. 2010. On the origin of shared beliefs (and corporate culture). The RAND Journal of Economics 41 (4): 617–648. Vaughan, A., and D. Rushe. 2017. Starbucks vows to hire 10,000 refugees as US companies condemn Trump travel ban. The Guardian, January 30. Welch, K., and A. Yoon. 2022. Do high-ability managers choose ESG projects that create shareholder value? Evidence from employee opinions. Review of Accounting Studies: 1–28. Wowak, A. J., J. R. Busenbark, and D. C. Hambrick. 2022. How do employees react when their CEO speaks out? Intra-and extra-firm implications of CEO sociopolitical activism. Administrative Science Quarterly 67 (2): 553–593. 40 APPENDIX A: VARIABLE DEFINITIONS Variable Note Activism variables [Source: Hand-collected from Factiva and X] 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [𝑘 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚_𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 [𝑘 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚_𝑝𝑜𝑙𝑖𝑡𝑖𝑐𝑠 [𝑘 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚_𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡 [𝑘 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚_𝑠𝑜𝑐𝑖𝑎𝑙 [𝑘 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚_𝑖𝑛𝑡𝑙 [𝑘 𝑚𝑜𝑛𝑡ℎ𝑠]𝑓,𝑚,𝑦 𝐴𝑐𝑡𝑖𝑣𝑖𝑠𝑚 [𝑦𝑒𝑎𝑟𝑙𝑦]𝑓,𝑦 A binary variable set to one for firm-months that are within k months after the CEO of firm 𝑓 engaged in CEO activism and zero otherwise A binary variable set to one for firm-months that are within k months after the CEO of firm 𝑓 engaged in CEO activism related to diversity and zero otherwise A binary variable set to one for firm-months that are within k months after the CEO of firm 𝑓 engaged in CEO activism related to political issues and zero otherwise A binary variable set to one for firm-months that are within k months after the CEO of firm 𝑓 engaged in CEO activism related to environmental issues and zero otherwise A binary variable set to one for firm-months that are within k months after the CEO of firm 𝑓 engaged in CEO activism related to social issues and zero otherwise A binary variable set to one for firm-months that are within k months after the CEO of firm 𝑓 engaged in CEO activism related to international relations and zero otherwise A binary variable set to one if the CEO of firm 𝑓 engaged in CEO activism in fiscal year 𝑦 and zero otherwise Employee sentiment variables [Source: Revelio Labs] 𝑀𝑒𝑎𝑛 𝑟𝑎𝑡𝑖𝑛𝑔 𝑚𝑎𝑛𝑎𝑔𝑒𝑚𝑒𝑛𝑡𝑓,𝑚,𝑦 𝑀𝑒𝑎𝑛 𝑟𝑎𝑡𝑖𝑛𝑔 𝑐𝑢𝑙𝑡𝑢𝑟𝑒𝑓,𝑚,𝑦 𝑀𝑒𝑎𝑛 𝑟𝑎𝑡𝑖𝑛𝑔 𝑜𝑣𝑒𝑟𝑎𝑙𝑙𝑓,𝑚,𝑦 𝑀𝑒𝑎𝑛 𝑟𝑎𝑡𝑖𝑛𝑔 𝑐𝑎𝑟𝑒𝑒𝑟𝑓,𝑚,𝑦 𝑀𝑒𝑎𝑛 𝑟𝑎𝑡𝑖𝑛𝑔 𝑏𝑎𝑙𝑎𝑛𝑐𝑒𝑓,𝑚,𝑦 Mean rating of senior management for firm 𝑓 in month 𝑚 of year 𝑓 Mean rating of culture and values for firm 𝑓 in month 𝑚 of year 𝑓 Mean overall rating for firm 𝑓 in month 𝑚 of year 𝑓 Mean rating of career opportunities for firm 𝑓 in month 𝑚 of year 𝑓 Mean rating of work-life balance for firm 𝑓 in month 𝑚 of year 𝑓 41 APPENDIX A (CONT’D) 𝑀𝑒𝑎𝑛 𝑟𝑎𝑡𝑖𝑛𝑔 𝑏𝑢𝑠𝑖𝑛𝑒𝑠𝑠𝑓,𝑚,𝑦 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡𝑓,𝑚,𝑦 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑝𝑟𝑜𝑠_𝑜𝑣𝑒𝑟𝑎𝑙𝑙𝑓,𝑚,𝑦 (𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑐𝑜𝑛𝑠_𝑜𝑣𝑒𝑟𝑎𝑙𝑙𝑓,𝑚,𝑦) 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑝𝑟𝑜𝑠_𝑚𝑓,𝑚,𝑦 (𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑐𝑜𝑛𝑠_𝑚𝑓,𝑚,𝑦) 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑝𝑟𝑜𝑠_𝑚𝑓,𝑚,𝑦 (𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑐𝑜𝑛𝑠_𝑚𝑓,𝑚,𝑦) 𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑝𝑟𝑜𝑠_𝑚𝑓,𝑚,𝑦 (𝐸𝑚𝑝_𝑠𝑒𝑛𝑡_𝑐𝑜𝑛𝑠_𝑚𝑓,𝑚,𝑦) Mean rating of business outlook for firm 𝑓 in month 𝑚 of year 𝑓 Moving average of overall rating for firm 𝑓 during the six-month window from year-month 𝑚 − 6 to 𝑚 − 1 standardized within each Fama-French 48 industry-year Decile ranking of the moving average of the proportion of employee ratings with overall rating of five (one) during the six-month window from year-month 𝑚 − 6 to 𝑚 − 1 Decile ranking of the proportion of employee reviews that mention the top management as pros (cons) of the firm during the six-month window from year-month 𝑚 − 6 to 𝑚 − 1 Decile ranking of the proportion of employee reviews that mention the corporate culture as pros (cons) of the firm during the six-month window from year-month 𝑚 − 6 to 𝑚 − 1 Decile ranking of the proportion of employee reviews that mention the politics as pros (cons) of the firm during the six-month window from year- month 𝑚 − 6 to 𝑚 − 1 Employee turnover variables [Source: Revelio Labs] 𝐿𝑜𝑔(𝑡𝑢𝑟𝑛𝑜𝑣𝑒𝑟 𝑐𝑜𝑢𝑛𝑡)𝑓,𝑚,𝑦 𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟 𝑟𝑎𝑡𝑒𝑓,𝑚,𝑦 𝐿𝑜𝑔(𝑟𝑒𝑐𝑟𝑢𝑖𝑡 𝑐𝑜𝑢𝑛𝑡)𝑓,𝑚,𝑦 𝑅𝑒𝑐𝑟𝑢𝑖𝑡 𝑟𝑎𝑡𝑒𝑓,𝑚,𝑦 𝐿𝑜𝑔(𝑅𝑒𝑐𝑟𝑢𝑖𝑡: 𝐻 𝑜𝑟 𝐴)𝑓,𝑚,𝑦 Natural log of one plus the number of U.S.-based employees that left firm 𝑓 in month 𝑚 of year 𝑓 The number of U.S.-based employees that left firm 𝑓 in month 𝑚 of year 𝑓 multiplied by 100 and divided by the average number of firm 𝑓 ’s employees in the past 12 months Natural log of one plus the number of all U.S.- based employees that joined firm 𝑓 in month 𝑚 of year 𝑓 The number of U.S.-based employees that joined firm 𝑓 in month 𝑚 of year 𝑓 multiplied by 100 and divided by the average number of firm 𝑓 ’s employees in the past 12 months Natural log of one plus the number of all U.S.- based employees who hold associate degrees or no college degrees that joined firm 𝑓 in month 𝑚 of year 𝑓 42 APPENDIX A (CONT’D) 𝐿𝑜𝑔(𝑅𝑒𝑐𝑟𝑢𝑖𝑡: 𝐵)𝑓,𝑚,𝑦 𝐿𝑜𝑔(𝑅𝑒𝑐𝑟𝑢𝑖𝑡: 𝑀 𝑜𝑟 𝐷)𝑓,𝑚,𝑦 𝐿𝑜𝑔(𝑆𝑎𝑙𝑒𝑠 𝑝𝑒𝑟 𝑒𝑚𝑝)𝑓,𝑦 𝑅𝑂𝐴𝑓,𝑦 Natural log of one plus the number of all U.S.- based employees who hold bachelor’s degrees but not graduate degrees that joined firm 𝑓 in month 𝑚 of year 𝑓 Natural log of one plus the number of all U.S.- based employees who hold graduate degrees that joined firm 𝑓 in month 𝑚 of year 𝑓 Natural log of firm 𝑓’s sales (in millions) in fiscal year 𝑦 divided by natural log of the number of employees (in thousands) at the end of fiscal year 𝑦 − 1 Firm 𝑓 ’s income before extraordinary items divided by total assets in fiscal year 𝑦 Control variables [Source: Compustat fundamental, CRSP] 𝑆𝑖𝑧𝑒𝑓[𝑞,𝑦] 𝑀𝐵𝑓[𝑞,𝑦] 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑓[𝑞,𝑦] 𝐶𝑎𝑠ℎ𝑓[𝑞,𝑦] 𝐵𝐻𝑅𝑓,𝑑 𝑅𝑒𝑡𝑉𝑜𝑙𝑓,𝑑 𝐿𝑜𝑔(𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠)𝑓,𝑚,𝑦 Natural log of firm 𝑓’s total assets (in millions) at the end of fiscal year 𝑦 or quarter 𝑞 Ratio of firm 𝑓’s market value of assets to book value of assets at the end of fiscal year 𝑦 or quarter 𝑞 Firm 𝑓 ’s debt divided by the sum of debt and common/ordinary equity at the end of fiscal year 𝑦 or quarter 𝑞 Firm 𝑓’s cash holdings divided by total assets at the end of fiscal year 𝑦 or quarter 𝑞 Firm 𝑓’s buy-and-hold stock market return for 250 trading days prior to day 𝑑 Standard deviation of firm 𝑓’s stock market return for 250 trading days prior to day 𝑑 Natural log of one plus the number of employees at firm 𝑓 at the beginning of month 𝑚 of year 𝑓 43 APPENDIX B: EXAMPLES OF CEO ACTIVISM Table B1: Examples of CEO Activism Date Firm CEO Source 2/27/2012 Berkshire Hathaway Warren E. Buffett AP Newswires 12/17/2013 Apple Inc Timothy D. Cook Washington Post 3/2/2018 4/5/2018 Campbell Soup Co JP Morgan Chase & Co Denise Morrison Philadelphia Business Journal James Dimon Reuters News 5/10/2019 Cabot Corp Sean Keohane ENP Newswire 6/4/2020 Best Buy Corie Sue Barry Warren's Consumer Electronics Daily 1/25/2021 Ecolab Inc Christophe Beck Twitter 9/2/2021 Match Group Inc Sharmistha Dubey Business Insider 44 Article title or tweet Billionaire Warren Buffett says wealthy should pay higher taxes, system should be reformed Apple CEO Tim Cook's candid comments about equality ; The Apple CEO speaks his mind about immigration reform and gay rights Campbell CEO Denise Morrison: It's time to 'shatter' glass ceiling JP Morgan's Dimon praises Trump's deregulatory efforts in annual letter Cabot Corporation CEO Sean Keohane Signs the CEO Action for Diversity and Inclusion Pledge Best Buy CEO Vows to Tackle 'Racial Injustice,' Pursue 'Systemic' Change Climate change requires a global response. On Wednesday, the U.S. took steps to rejoin the × ParisAgreement largest international effort to curb global warming. I'm encouraged by this move for the sake of our planet and for economic prosperity. Match Group CEO calls out Texas' 'regressive' anti-abortion law and says she will start a fund for employees who seek healthcare out of state the - Keywords taxes gay, healthcare Glass ceiling Immigration, Donald Trump inclusion racial Climate change Healthcare, abortion Category Diversity Environment Politics Social International relations Other APPENDIX C: CATEGORIES OF ACTIVISM BY TOPIC Table C1: Categories of Activism by Topic Keywords (case-insensitive) religion, lgbt, racism, same-sex, georgia house bill 757, women equality, equality act, transgender, pride month, blm, pride week, police brutality, homosexual, north carolina hb2, black lives matter, marriage equality, religious freedom act, gender equality, racial, bathroom bill, inclusion, equal pay, lesbian, glass ceiling, discrimination, gay, ethnicity, sexual discrimination bill, sexual harassment, harassment clean water, carbon tax, land conservation, paris accord, clean air, pollution, global warming, climate change, renewable, environment, to environment democrat, clinton, republican, trump, fiscal cliff, confederate statue, obama, confederate flag, nsa tracking, nsa data collection, taxes, government shutdown, debt ceiling, national security agency income inequality, human rights, poverty, white supremacists, censorship, planned parenthood, minimum wage, immigrants, healthcare, immigration, travel ban, indigenous people, refugee, terrorism, pay gap, universal healthcare, abortion, #keepfamiliestogether, dreamers impacts illegal war, tariffs, border security, brexit, nafta gun, mass shooting, violence, nazi, second amendment, charlottesville, prison, sanctions, advocate 45 APPENDIX D: EXAMPLE OF IDENTIFICATION IN THE FIRM-MONTH PANEL Table D1: Horizon of Activism Variables Date CEO activism 7/31/2019 8/31/2019 9/30/2019 10/31/2019 11/30/2019 12/31/2019 1/31/2020 2/29/2020 3/31/2020 4/30/2020 5/31/2020 6/30/2020 7/31/2020 8/31/2020 9/30/2020 10/31/2020 11/30/2020 12/31/2020 1/31/2021 2/28/2021 ✓ ✓ ✓ Activism [1 month] 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 Activism [6 months] 0 0 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 This table illustrates the identification strategy used for the firm-month panel. The checkmark denotes an occurrence of CEO activism during the month. Activism [1 month] (Activism [6 months])is set to one for firm-months that are within one month (six months) after the firm’s CEO engaged in activism and zero otherwise. 46 APPENDIX E: DYNAMIC ANALYSES Table E1: Comparison of Employee Sentiment Surrounding CEO Activism (1) Mean_rating_ management (2) Mean_rating_ management (3) Mean_rating_ culture (4) Mean_rating_ culture (5) Mean_rating_ overall (6) Mean_rating_ overall -0.018 (-0.42) -0.070 (-1.59) -0.045 (-1.07) -0.053 (-1.24) 0.064 (1.36) -0.006 (-0.16) 0.002 (0.06) -0.080** (-1.96) -0.017 (-0.47) 0.031 (0.77) -0.047 (-1.10) -0.080* -0.068 (-1.59) 0.054 (1.12) -0.017 (-0.40) -0.006 (-0.15) -0.090** (-2.18) -0.031 (-0.84) 0.016 (0.37) -0.061 (-1.45) 0.048 (1.12) -0.019 (-0.42) 0.030 (0.76) -0.071* (-1.77) -0.064* (-1.72) 0.052 (1.16) 47 -0.013 (-0.29) -0.042 (-1.07) 0.030 (0.70) -0.056 (-1.35) 0.055 (1.28) -0.015 (-0.33) 0.037 (0.95) -0.062 (-1.58) -0.054 (-1.45) 0.070 (1.61) -0.083** (-2.20) -0.047 -0.020 (-0.69) 0.027 (0.96) -0.014 (-0.40) -0.009 (-0.31) -0.051 (-1.61) -0.054* (-1.89) 0.063* (1.86) 0.010 (0.43) -0.028 (-0.99) -0.004 (-0.12) -0.018 (-0.60) 0.030 (1.09) -0.013 (-0.37) -0.007 (-0.23) -0.048 (-1.56) -0.053* (-1.87) 0.068** (2.00) -0.026 (-0.90) 0.008 VARIABLES Activism (m-6) Activism (m-5) Activism (m-4) Activism (m-3) Activism (m-2) Activism (m-1) Activism Activism (m+1) Activism (m+2) Activism (m+3) Activism (m+4) Activism (m+5) Table E1 (cont’d) VARIABLES (1) Mean_rating_ management Activism (m+6) Constant Observations Adj R-sq Controls Fixed effects SE clustered at 2.792*** (11.33) 53,960 0.10 Yes Firm, Year, Month Firm (2) Mean_rating_ management (-1.84) -0.012 (-0.28) 2.783*** (11.34) 53,740 0.10 Yes (3) Mean_rating_ culture 3.042*** (11.88) 53,960 0.19 Yes (4) Mean_rating_ culture (-1.03) -0.040 (-0.90) 3.046*** (11.90) 53,740 0.19 Yes (5) Mean_rating_ overall 3.464*** (17.33) 53,960 0.16 Yes Firm, Year, Month Firm, Year, Month Firm, Year, Month Firm, Year, Month Firm Firm Firm Firm (6) Mean_rating_ overall (0.29) -0.012 (-0.44) 3.480*** (17.34) 53,740 0.16 Yes Firm, Year, Month Firm 48 Table E2: Comparison of Employee Turnover Surrounding CEO Activism VARIABLES Activism (m-6) Activism × Mean_rating_overall (m-6) Activism (m-5) Activism × Mean_rating_overall (m-5) Activism (m-4) Activism × Mean_rating_overall (m-4) Activism (m-3) Activism × Mean_rating_overall (m-3) Activism (m-2) Activism × Mean_rating_overall (m-2) Activism (m-1) Activism × Mean_rating_overall (m-1) Activism (m) (1) (2) Log(turnover_ count) Log(turnover_ count) (3) Turnover_ rate (4) Turnover_ rate 0.027 (0.21) -0.003 (-0.08) 0.221* (1.66) -0.057* (-1.65) 0.202* (1.82) -0.061** (-2.20) 0.062 (0.47) 0.029 (0.23) -0.209 (-1.00) 0.005 (0.02) -0.000 (-0.01) 0.547** (2.04) -0.143** (-2.11) 0.180 (0.98) -0.056 (-1.20) -0.094 (-0.41) -0.011 -0.018 0.055 0.026 (-0.35) -0.005 (-0.05) (-0.51) -0.091 (-0.88) (0.99) -0.078 (-0.49) (0.43) -0.216 (-1.27) -0.006 0.015 0.015 0.050 (-0.23) 0.045 (0.35) (0.57) -0.039 (-0.28) (0.39) 0.033 (0.18) (1.24) -0.015 (-0.07) -0.017 0.002 -0.013 -0.003 (-0.53) 0.185** (2.00) (0.05) 0.165 (1.44) (-0.28) 0.272 (1.49) (-0.06) 0.317 (1.28) 49 VARIABLES Activism × Mean_rating_overall (m) Activism (m+1) Activism × Mean_rating_overall (m+1) Activism (m+2) Activism × Mean_rating_overall (m+2) Activism (m+3) Activism × Mean_rating_overall (m+3) Activism (m+4) Activism × Mean_rating_overall (m+4) Activism (m+5) Activism × Mean_rating_overall (m+5) Activism (m+6) Activism × Mean_rating_overall (m+6) Table E2 (cont’d) (1) Log(turnover_ count) (2) Log(turnover_ count) (3) Turnover_ rate (4) Turnover_ rate -0.045* -0.039 -0.063 -0.073 (-1.85) 0.002 (0.01) (-1.28) -0.119 (-1.07) (-1.34) -0.090 (-0.48) (-1.15) -0.201 (-0.95) 0.002 0.033 0.035 0.064 (0.06) 0.011 (0.08) (1.13) -0.081 (-0.62) (0.72) 0.043 (0.20) (1.17) -0.034 (-0.14) -0.001 0.024 -0.003 0.019 (-0.02) 0.288** (2.28) (0.70) 0.273** (2.09) (-0.06) 0.337* (1.65) (0.31) 0.286 (1.29) -0.077** -0.076** -0.076 -0.068 (-1.47) (-1.23) 0.535 (1.15) -0.133 (-1.12) 0.286 (1.32) -0.065 (-1.14) -0.037 (-0.19) 0.007 (0.14) (-2.36) (-2.27) 0.220 (1.20) -0.055 (-1.18) 0.102 (0.84) -0.028 (-0.88) 0.014 (0.11) -0.009 (-0.28) 50 Table E2 (cont’d) (1) Log(turnover_ count) -8.764*** (-13.81) (2) Log(turnover_ count) -8.804*** (-11.73) (3) Turnover_ rate 0.780** (2.15) 25,337 0.93 Yes Firm, Year, Month Firm 18,121 0.94 Yes Firm, Year, Month Firm 25,337 0.58 Yes Firm, Year, Month Firm (4) Turnover_ rate 0.867** (1.98) 18,121 0.61 Yes Firm, Year, Month Firm VARIABLES Constant Observations Adjusted R-squared Controls Fixed effects SE clustered at This table reports the results of OLS estimations that regress employees’ ratings of their firms (Panel A) and employee turnover (Panel B) on CEO activism at the firm-month level. Activism (m-k) [Activism (m+k)] are set to one for firm-months that are k months before [after] the firm’s CEO engaged in activism and zero otherwise. All variables are defined in Appendix A. Robust t-statistics are presented in parentheses. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 51 FIGURES Figure 1: Cumulative Abnormal Returns following CEO Activism This figure depicts the daily abnormal returns of firms whose CEOs engaged in CEO activism ("Activism firms") and other firms in the sample ("Control firms") for 11 days surrounding the day on which CEO activism occurs. Event day 0 is the day on which a) traditional media reports about the CEO of an activism firm expressing personal views on sociopolitical subjects or b) the CEO of an activism firm posts tweets with such views. I calculate abnormal returns using the market model with a pre-event estimation window of 250 trading days ending 30 days prior to the event day. 52 Figure 2.1: Timeline of Analyses for the Effects of CEO Activism on Employee Ratings Figure 2.2: Timeline of Analyses for the Effects of CEO Activism on the Sensitivity of Employee Turnover to Employee Sensitivity Figure 2.3: Timeline for the Effects of CEO Activism on the Firm’s Operating Performance This figure depicts the timeline of analyses in this paper. Panel A shows that I analyze the effect of CEO activism in month m on employee ratings in month m+1. Panel B shows that I analyze the effect of CEO activism and employee sentiment from month m to m+5 on employee turnover in month m+6. Panel C shows that I analyze the effect of CEO activism in year y on the firm’s operating performance in year y+1. 53 Tables Table 1.1: Sample Composition by Fiscal Year Fiscal year 2011 Firm-months 665 Firm-years 1,094 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Total 2,499 3,795 4,603 5,443 5,811 5,774 5,639 5,696 6,230 8,014 4,994 1,118 1,153 1,167 1,214 1,263 1,293 1,302 1,311 1,292 1,184 N/A 59,163 13,391 Table 1.2: Sample Construction Process Sample selection criteria Observations in the intersection of Execucomp, Compustat fundamental, CRSP, and Revelio Labs universe from 2011 to 2022, with non- missing dependent variables Less observations with less than 50 U.S.-based employees identified Less observations with missing control variables Final sample No. of firm-months No. of firm-years 79,682 16,625 (372) (20,147) 59,163 (1,808) (1,426) 13,391 This table describes the construction process for the data used in this study. My sample consists of firms in the intersection of Execucomp, Compustat fundamental, CRSP, and Revelio Labs universe from 2011 to 2022. Panel A presents the sample composition by fiscal year. Panel B reports the sample selection criteria for the firm-month and the firm-year panels. 54 Table 2.1: CEO Activism by Year and Month Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total 2011 2012 2013 2014 2015 2016 0 0 1 2 3 3 0 0 1 4 0 2 2017 25 19 2018 2019 2020 7 5 9 2021 13 5 6 1 6 0 2 0 2 2 6 6 1 1 2 2 4 5 3 0 0 0 3 2 7 8 20 12 13 7 3 5 2 10 14 3 4 6 1 1 2 1 3 7 9 14 12 32 10 0 0 2 4 2 2 8 0 0 1 1 0 2 8 11 15 5 8 8 7 5 1 0 1 1 1 3 7 9 9 3 8 5 0 6 3 1 6 8 7 7 5 17 14 0 4 0 4 2 11 7 9 5 4 6 1 3 1 1 5 6 5 4 2 6 9 3 18 14 26 32 66 114 126 65 99 102 665 55 Table 2.2: CEO Activism by Industry Industry Frequency Percentage Percentage in Execucomp universe Consumer Nondurables Consumer Durables Manufacturing Oil, Gas, and Coal Extraction and Products Chemicals and Allied Products Business Equipment Telephone and Television Transmission Utilities Wholesale, Retail, and Some Services Healthcare, Medical Equipment, and Drugs Finance Other 38 65 29 25 13 5.71 9.77 4.36 3.76 1.95 202 30.38 18 13 39 35 87 101 2.71 1.95 5.86 5.26 13.08 15.19 4.97 2.60 9.78 4.56 3.23 15.87 2.16 3.74 10.26 9.11 22.22 11.49 56 Table 2.3: CEO Activism by Topic Topic Diversity Politics Environment Social International relations Other Count 159 230 103 127 68 51 Percentage 23.91 34.59 15.49 19.10 10.23 7.67 Table 2.4: CEO Activism by Frequency throughout the Sample Period Frequency of CEO activism 0 Number of firms 1,084 1 2 3 4 5 6 7 8 9 10 or more Total 65 20 11 6 7 2 1 3 4 17 1,220 This table shows the frequency of CEO activism. Panel A shows the frequency of CEO activism by fiscal year and month; Panel B shows the frequency of CEO activism by industry following Fama- French 12-industry classification; Panel C presents CEO activism by topics, as categorized in Appendix C; Panel D shows the number of firms by the frequency of CEO activism throughout the sample period. 57 Activism variables Activism [1 month] Activism_diversity [1 month] Activism_politics [1 month] Activism_environment [1 month] Activism_social [1 month] Activism_intl [1 month] Activism_other [1 month] Activism [6 months] Employee sentiment variables Mean_rating_management Mean_rating_culture Mean_rating_overall Mean_rating_career Mean_rating_balance Mean_rating_business Emp_sent Emp_sent_pros_overall Emp_sent_cons_overall Emp_sent_pros_m Emp_sent_cons_m Emp_sent_pros_c Emp_sent_cons_c Emp_sent_pros_p Emp_sent_cons_p Employee turnover variables Log(turnover_count) Table 3.1: Summary Statistics (Firm-month panel) N Mean SD Min p25 Median p75 Max 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 54,807 56,665 56,672 56,672 56,672 56,672 56,672 56,672 56,672 56,672 0.009 0.002 0.003 0.002 0.002 0.001 0.001 0.033 2.867 3.102 3.574 3.016 3.129 3.796 0 5.276 4.803 5.022 5.264 4.664 4.032 1.141 2.324 0.093 0.049 0.054 0.040 0.043 0.031 0.030 0.179 1.227 1.330 0.993 1.182 1.210 1.206 0.995 2.975 3.368 3.220 2.986 3.444 3.637 1.117 3.066 0 0 0 0 0 0 0 0 0 0 1 0 0 1 -6.324 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 2 2.333 3 2.333 2.444 3 -0.567 3 1 1 3 1 1 1 1 0 0 0 0 0 0 0 0 3 3.167 3.714 3 3.067 4 0.059 5 5 5 5 5 1 1 1 0 0 0 0 0 0 0 0 3.800 4 4.182 4 4 5 0.635 8 8 8 8 8 8 1 1 1 1 1 1 1 1 1 1 5 5 5 5 5 5 4.181 10 10 10 10 10 10 10 10 59,163 3.587 1.344 0 2.708 3.526 4.477 7.902 58 Turnover_rate (%) Log(recruit_count) Recruit_rate (%) Log(recruit: H or A) Log(recruit: B) Log(recruit: M or D) Control variables Size MB ROA Leverage Cash BHR RetVol Log(employees) N 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 59,163 Mean 1.545 3.695 1.826 1.353 2.782 2.198 8.771 2.219 0.014 0.438 0.126 0.164 0.022 7.873 Table 3.1 (cont’d) Min 0 0 0 0 0 0 4.282 0.571 -0.243 0 0.001 -0.915 0.007 4.159 p25 0.946 2.833 1.008 0.693 1.946 1.386 7.576 1.224 0.004 0.235 0.030 -0.083 0.015 7.047 Median 1.336 3.638 1.559 1.099 2.708 2.079 8.713 1.689 0.013 0.420 0.078 0.107 0.019 7.817 p75 1.911 4.595 2.334 2.079 3.638 2.996 9.879 2.588 0.024 0.600 0.174 0.320 0.027 8.706 Max 9.357 8.004 11.824 5.971 6.901 6.288 15.115 13.896 0.128 2.103 0.726 8.228 0.104 11.740 SD 0.928 1.354 1.201 1.174 1.323 1.295 1.669 1.579 0.023 0.293 0.133 0.467 0.011 1.261 59 Mean_rating_management Mean_rating_culture Mean_rating_overall Log(turnover_count) Turnover_rate Log(recruit_count) Recruit_rate Size MB ROA Leverage Cash BHR RetVol Log(employees) Table 3.2: Tests of Mean Differences Activism – never 2.862 3.087 3.555 3.360 1.560 3.471 1.844 8.415 2.210 0.014 0.427 0.122 0.169 0.023 7.629 Activism – at least once 2.882 3.152 3.639 4.362 1.492 4.455 1.768 9.986 2.248 0.013 0.475 0.138 0.150 0.020 8.704 Diff St Err t-stat p-value -0.021 -0.065 -0.084 -1.002 0.069 -0.985 0.076 -1.572 -0.038 0.001 -0.048 -0.015 0.018 0.003 -1.075 0.012 0.013 0.009 0.013 0.009 0.013 0.012 0.015 0.015 0 0.003 0.002 0.005 0 0.011 -1.70 -4.95 -8.65 -80 7.55 -77.75 6.45 -104.40 -2.45 2.20 -16.60 -12.05 3.90 24.55 -92.95 0.092 0 0 0 0 0 0 0 0.014 0.026 0 0 0 0 0 60 Variables (1) Activism [1 month] (2) Mean_rating_management (3) Mean_rating_culture (4) Mean_rating_overall (5) Log(turnover_count) (6) Turnover_rate (7) Size (8) MB (9) ROA (10) Leverage (11) Cash (12) BHR (13) RetVol Table 3.3: Correlation Matrix (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) 0.00 0.01* 0.02* 0.10* -0.01 0.12* 0.02* 0.00 0.03* 0.02* 0.01 -0.01* 0.79* 0.64* -0.01* -0.02* 0.01* 0.06* 0.03* -0.04* 0.04* 0.00 -0.06* 0.59* 0.02* 0.01 0.01* 0.07* 0.02* -0.03* 0.05* -0.01 -0.07* -0.01* -0.07* 0.07* 0.10* 0.05* 0.00 0.06* 0.04* 0.02* 0.37* 0.51* 0.03* 0.08* 0.13* -0.04* -0.03* -0.19* -0.19* 0.06* -0.03* -0.06* 0.08* -0.03* 0.05* -0.20* -0.01* 0.25* -0.25* -0.04* -0.28* 0.41* -0.09* 0.39* 0.21* -0.10* -0.15* 0.12* 0.18* -0.25* -0.29* -0.01* 0.14* 0.10* 0.11* 0.09* This table reports descriptive statistics for the key variables and control variables used in the firm-month panel. Panel A reports summary statistics. Panel B displays tests of mean differences between firms whose CEOs never engaged in CEO activism during the sample period (“Activism – never”) and firms whose CEOs did at least once (“Activism – at least once”). Panel C reports a pairwise Pearson correlation matrix for selected variables. * denotes a two- tailed significance level of 5%. All variables are defined in Appendix A. 61 Table 4: Stock Price Reactions following CEO Activism: Daily Abnormal Returns Event day -5 -4 -3 -2 -1 0 1 2 3 4 5 Control Activism -0.004 -0.017 -0.006 -0.002 -0.001 -0.003 -0.029 -0.029 -0.017 -0.037 -0.045 0.008 0.048 0.024 -0.019 0.043 -0.054 -0.140 -0.030 0.026 -0.060 0.061 Diff -0.012 -0.065 -0.030 0.018 -0.044 0.051 0.111 0.001 -0.043 0.022 -0.106 St Err 0.055 0.054 0.055 0.055 0.055 0.055 0.054 0.054 0.054 0.054 0.054 t-stat -0.20 -1.20 -0.55 0.30 -0.80 0.90 2.02** 0 -0.80 0.40 -1.95* This table presents the daily abnormal returns of firms whose CEOs engaged in CEO activism ("Activism firms") and other firms in the sample ("Control firms") for 11 days surrounding the day on which CEO activism occurs. Event day 0 is the day on which a) traditional media reports about the CEO of an activism firm expressing personal views on sociopolitical subjects or b) the CEO of an activism firm posts tweets with such views. I calculate abnormal returns using the market model with a pre-event estimation window of 250 trading days ending 30 days prior to event day 0. All returns are expressed in percentage points. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 62 VARIABLES Activism [1 month] Size MB ROA Leverage Cash BHR RetVol Constant Observations Adj R-sq Fixed effects SE clustered at Table 5.1: Overall Effects of CEO Activism on Employees’ Ratings of Firms (1) Mean_rating_ management (2) Mean_rating_ management (3) Mean_rating_ culture (4) Mean_rating_ culture (5) Mean_rating_ overall (6) Mean_rating_ overall -0.093** (-2.19) 2.877*** (7,395.30) 54,140 0.10 Firm, Year, Month Firm -0.094** (-2.24) 0.002 (0.07) 0.033*** (3.60) 0.439 (1.18) -0.029 (-0.54) 0.093 (0.89) 0.038** (2.45) -1.407 (-1.47) 2.806*** (11.36) 54,140 0.10 Firm, Year, Month Firm -0.073* (-1.77) 3.115*** (8,249.74) 54,140 0.19 Firm, Year, Month Firm 63 -0.072* (-1.77) 0.012 (0.44) 0.013 (1.40) -0.251 (-0.63) -0.066 (-1.20) 0.138 (1.22) 0.050*** (3.09) -3.517*** (-3.56) 3.061*** (11.98) 54,140 0.19 Firm, Year, Month Firm -0.054* (-1.64) 3.556*** (11,696.37) 54,140 0.15 Firm, Year, Month Firm -0.055* (-1.66) -0.002 (-0.11) 0.029*** (4.20) 0.577* (1.94) -0.072 (-1.60) 0.144* (1.65) 0.022* (1.91) 1.706** (2.33) 3.477*** (17.47) 54,140 0.16 Firm, Year, Month Firm Table 5.2: Effects of CEO Activism on Employees’ Ratings of Firms by Topic VARIABLES Mean_rating_management Mean_rating_culture Mean_rating_overall (1) (2) (3) Activism_diversity [1 month] Activism_politics [1 month] Activism_environment [1 month] Activism_social [1 month] Activism_intl [1 month] Activism_other [1 month] Constant 0.053 (0.90) -0.158** (-2.30) -0.100 (-0.87) -0.106 (-1.45) -0.095 (-0.85) 0.130 (1.52) 2.799*** (11.36) 0.087 (1.44) -0.145** (-2.05) -0.075 (-0.87) -0.103 (-1.27) -0.192 (-1.63) 0.189** (2.10) 3.053*** (11.96) -0.002 (-0.05) -0.137** (-2.33) 0.047 (0.63) -0.057 (-0.86) -0.009 (-0.09) 0.048 (0.88) 3.468*** (17.47) Observations Adjusted R-squared Controls Fixed effects SE clustered at 54,140 0.10 Yes Firm, Year, Month Firm 54,140 0.19 Yes Firm, Year, Month Firm 54,140 0.16 Yes Firm, Year, Month Firm 64 Table 5.2 (cont’d) This table reports the results of OLS estimations that regress employees’ rating of their firms on CEO activism at the firm-month level. Panel A shows overall effects of CEO activism on employees’ ratings. Panel B reports effects of CEO activism on employees’ ratings of by topics, as categorized in Appendix C. Mean_rating_management (Mean_rating_culture) [Mean_rating_overall] is the mean rating of senior management (mean rating of culture and values) [mean overall rating]. Activism [1 month] is set to one for the first firm-months following occurrences of CEO activism and zero otherwise. All variables are defined in Appendix A. Robust t-statistics are presented in parentheses. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 65 Table 6.1: Effects of CEO Activism on the Sensitivity of Employee Turnover to Overall Employee Sentiment (1) Log(turnover_ count) (2) Log(turnover_ count) (3) Turnover_ rate (4) Turnover_ rate 0.040** (2.26) 1.575*** (28.86) -0.018*** (-6.46) -0.048** (-2.37) -8.114*** (-23.73) 56,622 0.89 Yes Firm, Year, Month Firm 0.055 (1.47) 1.571*** (28.71) -0.004*** (-3.95) 0.004*** (5.41) -0.009** (-2.18) 0.006 (1.33) -8.088*** (-23.66) 0.048* (1.83) -0.017 (-0.29) -0.021*** (-4.81) -0.075*** (-2.59) 0.395* (1.88) -0.004*** (-2.88) 0.006*** (5.20) -0.005 (-0.85) 0.017** (2.19) 0.407* (1.95) 56,622 0.89 Yes Firm, Year, Month Firm 56,622 0.48 Yes Firm, Year, Month Firm 56,622 0.48 Yes Firm, Year, Month Firm VARIABLES Activism [6 months] Log(Employees) Emp_sent Activism [6 months] × Emp_sent Emp_sent_pros_overall Emp_sent_cons_overall Activism [6 months] × Emp_sent_pros_overall Activism [6 months] × Emp_sent_cons_overall Constant Observations Adjusted R-squared Controls Fixed effects SE clustered at 66 Table 6.2: Effects of CEO Activism on the Sensitivity of Employee Turnover to Employee Sentiment by Topic VARIABLES Activism [6 months] Log(Employees) Emp_sent_pros_topic Emp_sent_cons_topic Activism [6 months] × Emp_sent_pros_topic Activism [6 months] × Emp_sent_cons_topic Constant Observations Adjusted R-squared Controls Fixed effects SE clustered at Top management (2) (1) Turnover_ Log(turnover_ rate count) -0.017 0.014 (-0.32) (0.39) 1.571*** (28.52) -0.000 (-0.63) 0.004*** (4.90) -0.000 (-0.03) 0.006*** (4.93) -0.006 (-1.16) 0.009* (1.72) -8.110*** (-23.52) 56,629 0.89 Yes Firm, Year, Month Firm -0.002 (-0.26) 0.012 (1.52) 0.366* (1.74) 56,629 0.48 Yes Firm, Year, Month Firm Culture Politics (3) Log(turnover_ count) -0.014 (-0.45) 1.569*** (28.42) 0.000 (0.22) 0.003*** (3.90) -0.002 (-0.43) (4) Turnover_ rate -0.073 (-1.39) 0.003** (2.01) 0.004*** (3.60) 0.004 (0.58) (5) Log(turnover_ count) 0.003 (0.12) 1.569*** (28.43) 0.003 (1.04) 0.003*** (2.72) -0.004 (-0.57) (6) Turnover_ rate -0.004 (-0.10) 0.006* (1.86) 0.005*** (4.04) -0.002 (-0.18) 0.010*** 0.016*** 0.008*** 0.010** (2.79) -8.070*** (-23.32) 56,629 0.89 Yes Firm, Year, Month Firm (2.62) 0.413** (1.96) 56,629 0.48 Yes Firm, Year, Month Firm (2.63) -8.080*** (-23.41) 56,629 0.89 Yes Firm, Year, Month Firm (2.22) 0.398* (1.89) 56,629 0.48 Yes Firm, Year, Month Firm 67 Table 6.2 (cont’d) This table reports the results of OLS estimations that regress firms’ employee turnover on CEO activism at the firm-month level. Panel A reports the effects of CEO activism on the sensitivity of employee turnover to overall employee sentiment. Panel B shows the effects of CEO activism on the sensitivity of employee turnover to employee sentiment by topic. Log(turnover_count) is the natural log of the number of U.S.-based employees that left the firm. Turnover_rate is the number of U.S.-based employees that left the firm multiplied by 100 and divided by the average number of employees in the past 12 months. Activism [6 months] is set to one for firm-months that are within six months after the firm’s CEO engaged in activism and zero otherwise. All variables are defined in Appendix A. Robust t-statistics are presented in parentheses. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 68 Table 7.1: Summary Statistics (Firm-year panel) Activism [yearly] Log(Sales per emp) ROA Size MB Leverage Cash BHR RetVol N 13,391 13,391 13,391 13,391 13,391 13,391 13,391 13,391 13,391 Mean 0.033 6.033 0.044 8.206 2.011 0.392 0.137 0.140 0.023 SD 0.179 0.871 0.069 1.659 1.324 0.270 0.143 0.351 0.011 Min 0 4.230 -0.217 4.895 0.810 0 0.002 -0.632 0.008 p25 0 5.452 0.011 6.983 1.138 0.187 0.030 -0.089 0.015 Median 0 5.920 0.041 8.110 1.552 0.383 0.086 0.112 0.02 p75 0 6.583 0.080 9.332 2.341 0.559 0.193 0.324 0.028 Max 1 8.633 0.242 12.112 9.090 1.139 0.596 1.528 0.080 69 Variables (1) Activism [yearly] (2) Log(Sales_per_emp) (3) ROA (4) Size (5) MB (6) Leverage (7) Cash (8) BHR (9) RetVol Table 7.2: Correlation Matrix (1) (2) (3) (4) (5) (6) (7) (8) 0.05* 0.02 0.22* 0.04* 0.07* 0.02 0.00 -0.03* 0.02* 0.29* -0.08* 0.08* -0.07* 0.01 -0.08* 0.02* 0.43* -0.18* 0.09* 0.20* -0.35* -0.22* 0.35* -0.34* -0.03* -0.31* -0.12* 0.42* 0.31* -0.05* -0.36* -0.05* 0.05* 0.10* 0.17* -0.07* 70 Table 7.3: Effects of CEO Activism on Operating Performance VARIABLES Activism [yearly] Size MB ROA Leverage Cash BHR RetVol Constant Observations Adj R-sq Fixed effects SE clustered at (1) Log(Sales_per_emp) (t+1) (2) ROA (t+1) 0.030** (2.00) 0.000 (0.02) 0.031*** (4.26) 0.857*** (7.87) -0.048 (-1.13) 0.149*** (2.63) 0.056*** (5.65) 1.312** (2.02) 5.910*** (34.38) 13,360 0.93 Firm, year Firm 0.005** (2.02) -0.011*** (-4.81) 0.013*** (12.09) 0.244*** (12.55) 0.001 (0.24) 0.011 (1.06) 0.023*** (12.14) 0.035 (0.32) 0.090*** (4.63) 13,360 0.58 Firm, year Firm This table reports descriptive statistics for the firm-year panel and regression results that examine the effects of CEO activism on firm’s subsequent operating performance. Panel A reports summary statistics for the key variables and some control variables used in the firm-year panel. Panel B reports a pairwise Pearson correlation matrix for selected variables. * denotes a two-tailed significance level of 5%. Panel C reports the results of OLS estimations that regress firm’s operating performance on CEO activism. Log(Sales_per_emp) is the natural log of sales (in millions) divided by natural log of the number of employees (in thousands) at the beginning of the year. ROA is the income before extraordinary items divided by total assets. Activism [yearly] is a binary indicator set to one for firm-year observations whose CEO engaged in CEO activism during the fiscal year and zero otherwise. All variables are defined in Appendix A. Robust t-statistics are presented in parentheses. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 71 Table 8: Placebo Tests: Effects of CEO Activism on Employees’ Ratings Unrelated to Values and Beliefs (1) Mean_rating_ career (2) Mean_rating_ balance (3) Mean_rating_ business -0.041 (-0.99) 0.021 (0.80) 0.031*** (3.78) -0.103 (-0.29) -0.090* (-1.79) 0.029 (0.28) 0.032** (2.20) -2.745*** (-3.03) 2.864*** (11.92) -0.034 (-1.04) 0.001 (0.05) 0.005 (0.54) -0.519 (-1.51) 0.031 (0.65) 0.132 (1.31) 0.029** (1.96) -1.805* (-1.91) 3.140*** (14.42) -0.045 (-1.00) 0.013 (0.41) 0.070*** (6.75) 2.494*** (6.51) -0.089 (-1.53) 0.082 (0.68) 0.156*** (9.51) -3.944*** (-3.74) 3.563*** (12.57) VARIABLES Activism [1 month] Size MB ROA Leverage Cash BHR RetVol Constant Observations Adj R-sq Fixed effects SE clustered at 54,140 0.09 Firm, Year, Month Firm 54,140 0.14 Firm, Year, Month Firm 50,274 0.14 Firm, Year, Month Firm This table reports the results of OLS estimations that regress employees’ ratings of their firms on CEO activism at the firm-month level. Mean_rating_career (Mean_rating_balance) [Mean_rating_business] is the mean rating of career opportunities (work-life balance) [business outlook]. Activism [1 month] is set to one for the first firm-months following occurrences of CEO activism and zero otherwise. All variables are defined in Appendix A. Robust t-statistics are presented in parentheses. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 72 Table 9.1: Effects of CEO Activism on the Sensitivity of Overall Employee Recruits to Employee Sentiment VARIABLES Activism [6 months] Log(Employees) Emp_sent Activism [6 months] × Emp_sent Emp_sent_pros_overall Emp_sent_cons_overall Activism [6 months] × Emp_sent_pros_overall Activism [6 months] × Emp_sent_cons_overall Constant Observations Adj R-sq Fixed effects SE clustered at (1) Log(recruit_ count) (2) Log(recruit_ count) (3) (4) Recruit_rate Recruit_rate 0.006 (0.29) 1.165*** (20.08) 0.004 (1.02) -0.022 (-0.77) -5.846*** (-14.86) 56,620 0.87 Firm, Year, Month Firm 0.115** (2.10) 1.162*** (19.96) 0.005*** (4.16) 0.003** (2.28) -0.013* (-1.69) -0.007 (-1.25) -5.867*** (-15.01) 56,629 0.87 Firm, Year, Month Firm 0.006 (0.17) 0.008 (1.14) 0.023 (0.59) 1.202*** (2.90) 56,620 0.52 Firm, Year, Month Firm 0.186** (2.01) 0.009*** (3.88) 0.004** (2.17) -0.010 (-1.08) -0.023** (-1.99) 1.139*** (2.79) 56,629 0.52 Firm, Year, Month Firm 73 Table 9.2: Effects of CEO Activism on the Sensitivity of Employee Recruits to Employee Sentiment by Education Levels VARIABLES Activism [6 months] Log(Employees) Emp_sent Activism [6 months] × Emp_sent Emp_sent_pros_overall Emp_sent_cons_overall Activism [6 months] × Emp_sent_pros_overall Activism [6 months] × Emp_sent_cons_overall Constant (1) Log(recruit: H or A) (2) Log(recruit: H or A) (3) (4) Log(recruit: B) Log(recruit: B) (5) Log(recruit: M or D) (6) Log(recruit: M or D) 0.016 (0.71) 0.830*** (16.05) 0.004 (1.20) -0.033 (-1.15) -5.266*** (-15.65) 0.115 (1.61) 0.829*** (16.01) 0.003*** (2.83) 0.000 (0.06) -0.015* (-1.70) -0.003 (-0.55) -5.282*** (-15.73) 0.002 (0.09) 1.138*** (20.39) 0.007* (1.96) -0.008 (-0.30) -6.476*** (-17.54) 74 0.133** (2.35) 1.134*** (20.22) 0.007*** (5.73) 0.003** (2.22) -0.013* (-1.65) -0.011** (-2.07) -6.503*** (-17.70) 0.005 (0.25) 1.005*** (18.12) 0.005 (1.34) -0.007 (-0.26) -6.292*** (-17.74) 0.166*** (2.89) 1.003*** (18.07) 0.005*** (3.73) 0.002 (1.64) -0.015* (-1.91) -0.015*** (-2.64) -6.315*** (-17.90) Table 9.2 (cont’d) VARIABLES Observations Adjusted R-squared Controls Fixed effects SE clustered at (1) Log(recruit: H or A) (2) Log(recruit: H or A) (3) (4) Log(recruit: B) Log(recruit: B) (5) Log(recruit: M or D) (6) Log(recruit: M or D) 56,620 0.80 Yes 56,629 0.80 Yes 56,620 0.84 Yes 56,629 0.84 Yes 56,620 0.83 Yes 56,629 0.83 Yes Firm, Year, Month Firm, Year, Month Firm, Year, Month Firm, Year, Month Firm, Year, Month Firm, Year, Month Firm Firm Firm Firm Firm Firm This table reports the results of OLS estimations that regress firms’ recruiting on CEO activism at the firm-month level. Panel A shows the effects of CEO activism on the sensitivity of overall employee recruits to employee sentiment. Log(recruit_count) is the natural log of the number of all U.S.-based employees that joined the firm. Recruit_rate is the number of U.S.-based employees that joined the firm multiplied by 100 and divided by the average number of employees in the past 12 months. Panel B reports the effects of CEO activism on the sensitivity of employee recruits to employee sentiment by employees’ levels of education. Log(recruit: H or A) {Log(recruit: B)} [Log(recruit: M or D)] is the natural log of the number of U.S.-based employees that joined the firm whose highest education is high school or an associate degree {a bachelor’s degree} [a master’s or doctoral degree]. Activism [6 months] is set to one for firm-months that are within six months after the firm’s CEO engaged in activism and zero otherwise. All variables are defined in Appendix A. Robust t- statistics are presented in parentheses. *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively. 75