Abstract
Violence that erupts in communities invites complex interpretations that can create a dilemma for business leaders about how to respond. Investigating how organizations respond to violence in protests, we build on the community embeddedness literature and propose that business leaders’ responses to protest violence depend on their perceptions of whether the violence is justified. Leaders may view reported protest violence as evidence of either social disorder or a valid grievance in the community where the violence occurs and where their companies are headquartered. We theorize that business leaders’ interpretations of violence are influenced by their community’s recent history: A history of protest violence unrelated to the social cause underlying the current protest weakens the managerial perception that the social issue is relevant to the community, while a history of grievance-validating events strengthens this perception. Using hand-collected data on corporate announcements following the 2020 Black Lives Matter (BLM) movement, we find that firms are less likely to announce diversity actions in response to reported protest violence in communities marked by persistent violence in previous non-BLM protests, but they are more likely to do so in communities with records of more police shootings. These same community conditions also shape the effect of violence on whether firms decide to publicly endorse the movement when they announce their corporate actions. Demonstrating that violence reshapes the ways that organizations respond to protests, we discuss the implications of our findings for research on violence, social movements, and corporate activism.
Keywords
In an increasingly interconnected and volatile world, the ways in which corporations interpret and respond to violence in their operating environments have profound implications for business and society. Acts of violence, whether rooted in political protests, individual outbursts, or the devastation of war, disrupt the social and economic fabric in which companies operate, requiring them to make complex decisions about altering business practices or responding in other ways (e.g., Hiatt & Sine, 2014; Østby, 2013; Sweet et al., 2024). Do firms align their actions with the causes behind the violence, potentially legitimizing or endorsing its underlying motivations? Or do they distance themselves, maintaining a neutral stance even when silence could be interpreted as dismissal? These questions probe the boundaries of corporate responsibility and strategy, highlighting growing tensions regarding shareholder interests and societal expectations in times of conflict. Understanding how businesses interpret and act on violence reveals much about their role as social actors in moments of upheaval.
Among the various forms of violence, protest violence presents a particularly useful lens through which to explore corporate behavior. Protest movements often challenge existing power structures, drawing attention to systemic inequalities or pressing social issues. When violence erupts during protests, whether through clashes with law enforcement, destruction of property, or confrontations between opposing groups, it can indirectly force companies into the fraught territory of deciding how to respond (Dimitriadis, 2021). Although business leaders may not feel obligated to respond directly to such protests, they might draw inferences about the changing relevance for the local community of the social issue being protested (Ingram et al., 2010). For example, building on an activism-as-information logic, Zhang et al. (2023) showed that the scale of street protests in communities influences whether locally headquartered companies act to better align their organizations with a movement’s goals.
Drawing on research from the field of social movements (Kim & Schifeling, 2022; King, 2008; King & Soule, 2007; Luo et al., 2016; Pacheco & Dean, 2015; Rao & Sivakumar, 1999), we examine how firms navigate these dynamics, focusing on whether their actions amplify, suppress, or remain detached from the causes championed in protests with violence. To do so, we consider how business leaders interpret and respond to reported violence in community protests resulting in physical damage to people or property.
Unlike other attributes of protests, violence invites complex and possibly conflicting interpretations. Other protest attributes, such as scale or frequency, typically trigger a straightforward stakeholder-inference logic: The more people who protest, or the more events they organize, the more likely firms are to favorably signal alignment with their stakeholders (e.g., Marquis & Bird, 2018; Zhang et al., 2023). In contrast, reported protest violence could prompt the opposite reaction. Corporate leaders may view violence as a sign of social instability and deliberately distance themselves to avoid being associated with what they perceive as societal disorder. Bailey and colleagues (2023), for instance, found that violence can undermine a movement’s credibility, while others have shown that it reduces public support (Agarwal et al., 2025; Thomas & Louis, 2014) and lowers the chances of target concessions (Rojas, 2006). If business leaders interpret protest violence negatively, they may be even more reluctant to take actions that could be construed as supporting the movement. However, some executives might, instead, interpret violence as evidence of the issue’s urgency and relevance, prompting them to adjust their business practices to align with the protest’s underlying cause.
Given this complexity and the potential for divergent interpretations of protest violence, we theorize that business leaders’ responses to local protest violence will hinge on their perceptions of whether the violence is justified. While assessing the justification for violence is inherently challenging, we argue that one dimension that most business leaders will account for is the historical conditions of the communities in which the protests occur and their companies are headquartered. This perspective aligns with a robust literature showing that business leaders often interpret and respond to societal events through a localized community lens (Hiatt et al., 2009; Lee & Lounsbury, 2015).
On this basis, adapting a distinction invoked by Turner (1969), we propose that the interpretation of protest violence as evidence of either social disorder or a valid grievance depends on a community’s recent history. A pattern of past protest violence in a community may lead observers to see new instances of violence as a continuation of social disorder, reducing the perceived legitimacy of the cause and discouraging business leaders from taking supportive action. But recent events that validate community grievances can frame protest violence as a more justified response to systemic issues, increasing its perceived local relevance. When business leaders and local stakeholders interpret the violence through this lens, local organizations are more likely to announce actions in support of the social cause underlying the protest. Moreover, for firms that decide to announce such actions, the same community context features will also shape how likely they are to endorse the grievance-related social movement in their announcements.
To test our theory, we examined announcements of corporate diversity actions during the 2020 Black Lives Matter (BLM) movement. BLM protests against racism proliferated across the United States in response to the death of George Floyd, a Black man who was killed when a police officer knelt on his neck. The BLM movement, one of the largest street protests in American history, garnered significant attention from business leaders and stakeholders even though it did not directly target businesses. We analyzed hand-collected data on weekly corporate diversity actions announced by public Fortune 500 companies from May 25, 2020 to December 27, 2020, in conjunction with data on street protests in the communities where these companies were headquartered.
Our findings provide considerable support for our theory. While peaceful protests were not associated with an increase in local corporate diversity action announcements (relative to periods without protests), protests with reported violence were positively associated with such announcements—under certain predicted community conditions. Specifically, communities with more reported protest violence reflected increased corporate diversity action announcements when they had a minimal history of violence in previous non-BLM protests and when they had a strong history of grievance-validating events. A community’s recent history also shaped the effect of violence on a firm’s decision to endorse the BLM movement while making a corporate diversity action announcement. In particular, we found that reported violence in BLM protests was associated with the public endorsement of the movement if the community lacked a history of protest violence from prior, non-BLM protests.
This study reveals how violence reshapes the ways in which business organizations respond to protests. Prior research has suggested that companies’ support for protested issues is largely driven by signals of stakeholder support, such as protest scale (e.g., Zhang et al., 2023). The prevailing logic is that the more stakeholders appear to care, the more likely firms are to align their actions with the cause. However, violence introduces a fundamentally different dynamic, placing business leaders in a dilemma: Should they support a cause associated with protest violence and risk being perceived as endorsing violence, or should they remain silent to distance themselves from the unrest, even at the cost of alienating stakeholders? Drawing on research on community embeddedness, we develop a community-centric framework to explain how business leaders interpret reported protest violence, shaping both their action announcements and their social movement endorsements.
These findings provide new insights at the intersection of several research streams. To the growing literature on how businesses respond to societal violence (Hiatt & Sine, 2014), our study offers a novel community-based decision framework. While this framework is anchored in findings specific to protest violence, we discuss how its underlying logic can be generalized to understand the impact of various other forms of violence on business. Additionally, our findings advance conversations on how social movements shape markets (King & Soule, 2007), by shifting from a unidirectional theory of activism-as-information—focused solely on protest features that provide relatively straightforward clues regarding the relevance of a social issue—to a more contingent perspective in which interpretations of protests depend on features of the local community. In light of the surge of interest in corporate activism in recent years (Hambrick & Wowak, 2021; McKean & King, 2024), our study also provides timely new insights into the factors that can compel companies to not only respond to contentious social movements but also to publicly endorse them in their corporate actions.
Organizational Responses to Protest Violence
How do organizations respond to protest violence? In considering this question, we start from the view that corporate responses to protests in a local community are likely to hinge on the inferences that business leaders and local stakeholders make as they learn about protest events.
Organizations’ Community Embeddedness
Companies are heavily embedded in the local communities where they are headquartered (Tilcsik & Marquis, 2013), and business leaders’ interpretations of and responses to societal events and trends are often filtered through a localized community lens (Hiatt et al., 2009; Lee & Lounsbury, 2015). Even in large multinationals, CEOs, their fellow executives, and a significant proportion of corporate stakeholders tend to live in or near the communities where their companies are headquartered (Duchin & Sosyura, 2021). As a result, these individuals likely observe and absorb signals from their local environment through conversations with community members, employees, and civic leaders, as well as through local events and news coverage.
Business leaders may infer the extent of local support for social causes based on these cues from their communities. For example, witnessing or hearing about a large local protest may lead a business leader to conclude that the underlying social issue is more relevant to the local community than previously thought (Ingram et al., 2010). Such inferences of cause relevance may, in turn, prompt business leaders to consider corporate actions that better align with the community’s apparent support for the protest cause, as Zhang and colleagues (2023) found for firms responding to the scale (i.e., number of supporters) of local Women’s March street protests.
Evidence shows that executives of large corporations are influenced by and strive to contribute to their local communities. Business leaders often “value the community for its own sake” (Almandoz et al., 2017, p. 192) and seek to make positive contributions to strengthen the sense of identity it provides for organizational members (Marquis et al., 2011). At the same time, they may embrace a more instrumental logic, since important stakeholders, including employees and suppliers, are likely to reside in and care about the community in which a firm is headquartered. Referencing both intrinsic and instrumental pathways, researchers have examined how the communities where firms are headquartered influence various types of decisions that leaders make, including those related to philanthropic contributions (Tilcsik & Marquis, 2013), environmental management practices (Kim et al., 2019; Lee & Lounsbury, 2015), and corporate social responsibility (CSR) activities (Husted et al., 2016; Sun et al., 2015).
When we apply this logic to local protests, if business leaders infer greater cause relevance in their community, they should be more likely to support the protest cause through corporate actions for both intrinsic and instrumental reasons. Intrinsically, business leaders may feel a moral motivation to support a protest cause that is relevant to the local community where they live and work and which they identify with and seek to support (Almandoz et al., 2017). Instrumentally, business leaders may also take action to maintain the support of aggrieved local groups or sympathetic stakeholders, who may be protest participants themselves. Business leaders may therefore engage in corporate actions that support the protest cause, assuming those actions can help their organizations gain and maintain community-based stakeholder support. This argument was supported by a global investment manager we interviewed, who explained that in the wake of BLM protests, “there was a sea change that happened in the way that management teams started looking at [trying] to attract and retain the best workforce, especially at the younger ages, which tends to be the people who care deeply about this.”
Violence in Protests
Yet, it remains unclear how business leaders interpret reported protest violence. Numerous studies have shown that violence can advance movement objectives, at least under certain conditions. Studying an outbreak of protest violence in Los Angeles in 1992, Enos et al. (2019) found that these protests increased local support for the goals of a movement seeking to address police brutality. While acknowledging that violence might provoke backlash elsewhere, the authors reasoned that protest violence could build sympathy among local observers. Underscoring the role of the local population, Shuman and colleagues (2022) proposed that violence that manifests within a larger nonviolent movement can increase support for a protest’s goals among skeptics who live in the area of the protest violence. In this vein, Fording (1997) found that protest violence can drive policy changes that are favorable to protesters, contingent on local democratic political institutions.
But studies also indicate that reported protest violence undermines outsiders’ support for a movement, mostly by undermining the movement’s perceived legitimacy (e.g., Agarwal et al., 2025; Feinberg et al., 2020; Steinert-Threlkeld et al., 2022). In the context of student activism, for instance, Rojas (2006, p. 2147) argued that disruptive protests involving violence “deprive administrators of the legitimacy needed to enact change,” thereby hampering organizational support for the movement. Similarly, Wasow (2020, p. 643) found that violent protests decrease public support because outsiders perceive them as “riots.”Bailey and colleagues (2023, p. 1208) suggested that violent movements might be “perceived as less worthy” than movements without violence. For business leaders, protest violence could either increase or decrease the perception of cause relevance in the local community, making it important for scholars to consider the community in which the violence is embedded.
Community History as a Lens Driving Contingent Responses
According to the activism-as-information logic (King & Soule, 2007; Zhang et al., 2023), protest features should, in general, provide business leaders with information about community sentiment regarding the protested social issue. However, while many protest attributes, such as scale and frequency, can be interpreted by business leaders and other observers as basic indicators of community sentiment regarding the protested social issue, we argue that the signal of violence operates differently. At one extreme, managers could perceive violence as evidence of people’s suffering and frustration over injustice, in other words, as partly justified; at the other extreme, they could perceive violence as being perpetrated by troublemakers who are not following the rules. Turner (1969) highlighted that violent U.S. civil rights protests during the 1960s were often interpreted in starkly different ways, as evidence of either social disorder or strongly felt valid grievances. These two interpretations have opposing implications: If violence is evidence of social disorder, business leaders should think the protest issue is less relevant to the community, while if violence is an expression of valid grievance, the issue is more relevant.
When confronted with protest violence, observers such as business leaders could infer, with varying degrees of emphasis, either social disorder or valid grievances. Crucially, we argue, this emphasis can be tilted one way or the other by the characteristics of the community where the protest violence occurs. In community-based research (Marquis et al., 2007), events in a community’s history that are relevant to local community members can serve as a proxy for the community’s unobservable characteristics, such as the level of trust in a community (Greve & Yue, 2017) or its pre-existing grievances. As local community members digest reports of protest violence, we argue, they are likely to interpret the meaning of that violence based on contextual information about their community’s recent past.
Recent history of protest violence
One component that is relevant to business leaders’ interpretations of violence reported in a focal protest is the community’s recent history of violence in other protests, especially those with different goals than those of the current protest. As street protest is a geographically bounded activity that occurs within a community, the traditions, political cultures, and economic conditions of the community shape how protests develop and play out there (Navickas, 2015; Rucht, 1998). For example, Bohstedt and Williams (1988, p. 21) found that a tradition of protest violence can be “part of the collective mentality of a given community,” making violence in future protests more likely. This community-centered history of protest violence can thus serve as a relevant contextual factor that can be observed by all community members and influence how protest violence is interpreted when a new protest emerges.
According to research on the “diversity effect” in category-based cognitive psychology (Hayes et al., 2019, p. 1043; Osherson et al., 1990), observers are more likely to generalize a given characteristic to a broader category when it applies to multiple different items within that category. In our context, if community members recall ongoing violence in other community protests on different social issues, they should generalize reported violence in a current protest to the broader category of that community’s protest tradition or tendency. Therefore, reported violence in a current protest is more likely to be seen as evidence of social disorder in communities that have recent histories of persistent protest violence, as it reflects an ongoing pattern of social disorder rather than a strongly felt grievance. The logical inference is thus that the protest cause has less relevance in the wider community than might otherwise be understood from the protest taking place. Additionally, business leaders may become desensitized to the novelty of violence as a result of its prevalence across different eras and issues, which could further diminish their responses to newly observed violence (Morgeson et al., 2015).
Moreover, in communities with a history of protest violence, members, including corporate leaders and employees, may have a pre-existing negative narrative about such violence. Having observed protest violence in their community in the past, community members may believe that it is not only unproductive but also disruptive to the community’s social order. Indeed, they will have been exposed to the negative consequences of protest violence in the past, such as property damage and emotional distress, without seeing the benefits of those efforts. In turn, when new protest violence occurs in their community, it is likely to be interpreted through the lens of these pre-existing beliefs. Influenced by their confirmation bias (Nickerson, 1998), community members interpret new violence as disruptive and merely part of the community’s protest tradition, rather than as a reflection of valid community grievances.
The interpretation of violence as evidence of social disorder decreases the extent to which the social movement’s cause is perceived as relevant to the community, thus reducing business leaders’ intrinsic and instrumental motivations to initiate corporate responses that support the causes driving protests in their communities. We thus anticipate that the effect of reported protest violence on local corporate actions that support the current social cause depends on the community’s history of protest violence: 1
After business leaders decide to support the cause associated with protest violence, they must determine whether to formally endorse the social movement when announcing their organizational actions. Organizations responding to a BLM protest, for example, could simply announce a new donation benefiting local Black residents or directly frame that initiative as support for the BLM movement. A response from Amazon.com to the BLM movement is an example of the latter. In an announcement about a $10 million donation to improve the lives of Black Americans, the company clearly stated, “We believe Black lives matter.” Publicly endorsing the movement by effectively crediting it for an organizational action has the effect of relationally affiliating the organization with the social activist movement (Kim & Schifeling, 2022).
Existing research indicates that this intertwining of reputations and fates creates additional opportunities and risks for the focal firm. First, while corporate ties with activists can reduce tension between the focal firm and other like-minded activists in the same broad movement (Odziemkowska & McDonnell, 2024), corporate endorsement of an activist group may provoke backlash from opposing activist groups and stakeholders who aim to defend the institutional order and penalize companies that support opposing movements (Bondi et al., 2025; McDonnell, 2016; Meyer & Staggenborg, 1996). These reactions could generate negative attention and accusations of hypocrisy for these companies (Jin et al., 2025; Li & Soule, 2021). Second, affiliating with a social movement increases a firm’s risk of being targeted by future activist campaigns. Because endorsing activists when announcing organizational actions can signal that the company and its leaders are sympathetic or responsive to stakeholder demands (Briscoe & Safford, 2008; Gupta & Briscoe, 2020), findings show that companies that concede to activists’ demands are more likely to attract additional attention and pressure from activists in the future (McDonnell et al., 2015).
While this existing theory indicates that affiliating with activists carries the risk of backlash from counter-activists and invites further activism, the decision to endorse a visibly violent protest movement brings a different type of risk: disapproval from broader stakeholders and public audiences. This is because such an endorsement publicly links the business not just with the cause but also, to some extent, with the reported acts of violence. Business leaders might anticipate such action-with-endorsement to be met with approval from stakeholders who share those views and with skepticism or opposition from others within the broader community and stakeholders (Bhagwat et al., 2020; Hurst, 2023). In contrast, announcing corporate actions without movement endorsement may still reflect the view that the protest violence was justified and reflected valid grievances, consistent with our arguments above, but stops short of directly associating the organization with violence or with groups that reportedly behaved violently. Maintaining this distance by announcing actions without movement endorsement could thus protect an organization from subsequent blowback. One executive we interviewed explained:
We would acknowledge our community is hurting and our Black neighbors are hurting. We would recognize we maybe need to think about how we make a charitable donation or lend some volunteerism to a cause that is related to this, but maybe not directly endorsing it. Something that’s helping our Black community in [our headquarters city]. So, related but maybe not direct.
Given this predicament that responsive organizations may encounter, and turning to our argument that a community’s recent history forms a primary lens through which business leaders interpret protest violence, we can consider how a community’s history of protest violence may influence corporate willingness to associate with activist groups involved in protest violence. Absent what is perceived to be a valid grievance, reports of protest violence should weaken an organization’s willingness to associate itself with a movement through endorsement. However, if a community’s recent history supports the inference that the reported protest violence reflects valid grievances rather than social disorder, business leaders may perceive a lower risk of backlash from local stakeholders and thus be more inclined to endorse the movement when announcing their corporate actions.
Specifically, in a community with more historical protest violence across different causes, the negative effect of reported protest violence on endorsement may be amplified because such a history undermines the inference that the violence stems from a valid grievance. Instead, the history supports the inference that the current episode of protest violence is part of the community’s tradition of social disorder. This inference makes movement endorsement appear less legitimate to community stakeholders, reducing business leaders’ intrinsic and instrumental motivations to publicly endorse the movement. Therefore, among firms that take actions to support the social cause driving protests, the effect of protest violence on endorsement of the movement depends on the community’s history of protest violence:
Recent history of grievance-validating events
Beyond a community’s history of violence in unrelated protests, the violence perceived in a local protest may be influenced by the degree to which the protested grievance has historical roots in the community where it occurs. The collective memory shared by community members (Della Porta et al., 2015), including business leaders and local stakeholders, can shape their interpretations of reported protest violence. For instance, one executive we interviewed shared the memory of recent community events that shaped their interpretations of racial unrest in Buffalo, New York:
We know in Buffalo right now there are some really tense and unfortunate relations between our Black neighbors and our White police officers—our Black community was tremendously affected by the May 14 [2022] shooting at Tops Markets, which was a racially motivated hate crime that killed 10 people.
In a community with a recent history of grievance-validating events, observers may be more understanding when violence occurs in current protests (Olzak, 2021; Opp, 1988). For instance, in a community where police have been violent toward and targeted minorities, violence among local minority protesters is more likely to be interpreted as an expression of a valid grievance, substantiated by the harm community members have observed from the historical record of policing. In contrast, violence in environmental protests would more likely be interpreted as evidence of social disorder in a community that has never experienced an environmental disaster, as community members would have less contextual evidence to justify protesters’ violent behavior.
When business leaders and local stakeholders interpret protest violence as reflecting a valid grievance, local companies are likely to perceive the social cause as highly relevant to the community, leading to increased intrinsic and instrumental motivations to announce corporate support. Our argument implies that a positive (negative) general response to protest violence is strengthened (weakened) in communities with more grievance-validating historical events. Therefore, the effect of reported protest violence on local corporate actions that support the social cause depends on the community’s history of grievance-validating events:
If we continue with this logic, a community’s history of grievance-validating events should also impact business leaders’ assessments of whether to endorse the cause underlying a violent protest. In general, protest violence that is perceived as lacking a valid cause should discourage business leaders from formally endorsing the associated social movement. However, if a community has a strong history of grievance-validating events, protest violence should have a weaker negative effect on corporate endorsement because business leaders are likely to infer that the violence legitimately reflects unjust suffering in the community. Hence, among firms that take actions to support the social cause driving protests, the effect of protest violence on endorsement of the movement depends on the community’s history of grievance-validating events:
Methods
Research Setting: The BLM Movement
Our theory addresses whether and under what conditions reported violence in local street protests prompts firms to take actions that align with activists’ goals and to publicly endorse social movements. We test this theory in the context of the BLM movement for two reasons. First, the movement’s core issues—diversity and inclusion—are broadly relevant across the business sector. Second, the movement’s protest events are well documented and, importantly, were not directed at specific firms, allowing us to observe organizational responses in the absence of targeted pressure.
BLM is a social and political movement that originated in the U.S. in July 2013 following the acquittal of an individual involved in the fatal shooting of Trayvon Martin, a Black teenager. The movement seeks to address systemic racism and violence against Black Americans, with particular attention to incidents involving law enforcement. In 2020, the BLM movement gained renewed national and international attention after the death of George Floyd on May 25. Video footage showed Floyd repeatedly stating “I can’t breathe” while being restrained by police. The incident sparked widespread public reaction and became a catalyst for a surge in protests and activism associated with the movement. Estimates from various polls suggest that between 6 and 10 percent of the U.S. population participated in at least one BLM protest in 2020 (Buchanan et al., 2020), marking it as one of the largest protest movements in American history.
Important for our analytical purposes, the BLM protests unfolded across many communities in the U.S. where different companies were headquartered. According to Duchin and Sosyura (2021), in 2020 more than 96 percent of CEOs lived within 100 miles of their corporate headquarters. Executives who live in the same community as their corporate headquarters are likely to have obtained information about local protests, including the BLM protests, from local or national newspapers, indirectly through their personal ties in the local community, or from other employees at corporate headquarters who reside in that community.
Within this context, our analysis considers the impact of violence reported in protests. Although most BLM protests have been peaceful, reports of violence during some protests received considerable attention from outsiders. While research shows that more than 90 percent of BLM protest events resulted in no injuries to police and no property damage (Adhikari, 2021), when violence did occur, it was often reported in news stories.
While developing our project, we conducted five interviews in an effort to better understand how business leaders respond to community events, including street protests and violence. The interview participants were all senior managers in general management or sustainability positions in large publicly traded U.S. corporations. These interviews represent a non-systematic convenience sample of relevant business leaders who were identified and contacted through a university alumni office. The semi-structured interviews lasted 30 to 40 minutes and followed an interview guide that started with the question, “What does your company account for when deciding which CSR initiatives to develop?” and then moved to more-specific queries regarding attention to social movements, protests, and other developments in the communities where companies are headquartered. Insights from these interviews informed our theorizing, and a few interview quotes are included in this article.
Data and Sample
To compile information on the BLM protests, we used data from the Crowd Counting Consortium (CCC), which collects publicly available data on marches, strikes, street protests, and other political crowds. This hand-coded dataset synthesizes information from multiple sources, including mass media, organization websites, police reports, search engines, and social media (Fisher et al., 2019). It has been used in recent sociology and political science research (Beyerlein et al., 2018; Smith et al., 2020), and its accuracy has been supported by other data sources (Sobolev et al., 2020). The dataset contains information about protests that have taken place in the U.S., including the dates and locations of events, the issues being protested, the estimated size of the crowd at each event, and any reported instances of violence. Leveraging these data, we identified 16,484 BLM protests and 904 counter-BLM protests in the U.S. from May 25 to December 27, 2020.
For our dependent variable, we hand-collected information from media reports and press releases to construct a unique dataset of corporate diversity action announcements, which we describe below. The remaining data came from other sources, including police shooting data from The Washington Post’s Fatal Force Database 2 ; community characteristics data from the U.S. Census Bureau and MIT’s Election Data and Science Lab; COVID-19 data from The New York Times 3 ; CEO political ideology data provided by Christensen et al. (2015); and firm financial data from Compustat.
We used core-based statistical areas (CBSAs) to link firms to the communities where they are headquartered and to define protests as local. CBSAs encompass workplaces and residential neighborhoods via common commuting routes. In line with previous research (Ferguson et al., 2018; Marquis et al., 2007), we captured community-related variables, including reported protest violence, at the CBSA level rather than the city level, because CBSAs more accurately capture economic and social connections within communities. Crosswalks from the National Bureau of Economic Research and the U.S. Department of Housing and Urban Development were used to connect the locations of corporate headquarters (by ZIP code) to the locations of protests (by Federal Information Processing Standards codes) and to aggregate the firm–protest data to the CBSA level. Figure 1 illustrates the extent of BLM protests in 2020 and the reported violence in those protests by CBSA. As shown, protests occurred in a wide range of communities with varying intensity, and peaceful protests were more common than protests in which violence was reported.
BLM Protests with Reported Violence vs. Peaceful BLM Protests, by CBSA (2020)*
Our sample comprises 14,508 firm–week observations from 468 publicly traded Fortune 500 firms in 2020. For our weekly panel, we defined the first weekday as Monday because most BLM protests happened on Saturdays and Sundays, which also allowed us to group both weekend days into the same week and to capture how local BLM protests occurring each weekend influenced firms’ announcements of actions the following week.
Our observation window begins on May 25, 2020, the Monday that George Floyd died, and runs through December 27, 2020, the Sunday of the last full week of that year. As shown in Figure 2, while there was a large wave of protest violence immediately after George Floyd’s death on May 25, there were several additional waves of protest violence thereafter in various communities across the U.S. For instance, in the New York-Newark-Jersey City area, in addition to the protest violence that occurred immediately after Floyd’s death (in weeks 22 and 23 of 2020), there were also reports of protest violence occurring in weeks 27, 30, 36, 41, and 44 of the year. We ended our observation period to capture these events while also avoiding potential effects of the U.S. Capitol attack on January 6, 2021. As we will show in Table A5 in the Online Appendix, our results are consistent when we ended our observation period on the Sunday right before the 2020 U.S. presidential election (i.e., November 1, 2020).
Number of BLM Protests with Reported Violence, Corporate Diversity Actions, and Movement Endorsements Over Time in Four CBSAs
Constructing a Dataset of Corporate Diversity Actions
We collected data on corporate diversity actions for all publicly traded Fortune 500 firms in 2020. Following social movement research (Dorobantu et al., 2017; Marquis & Bird, 2018; McDonnell & Cobb, 2020), we hired three research assistants (RAs) to read and code information from press releases and media articles from Factiva to construct our focal dependent variables. We included 11 media outlets (The New York Times, The Wall Street Journal, USA Today, CNN, The Washington Post, Chicago Tribune, Los Angeles Times, Fox News, CBS Network, NBC Network, and ABC Network) that represent diverse political orientations and geographic regions. We also coded actions based on the complete corpus of firms’ press releases on Factiva.
The three RAs manually read all press releases and media articles for every publicly traded Fortune 500 company that were yielded by a list of keywords, which is presented in Table A1. The method for developing the keyword list and more details of the data collection are provided in Online Appendix 1. The RAs evaluated whether each article described an announcement of a corporate diversity action, and coded each action accordingly. At the screening stage, we defined corporate diversity actions broadly as any new actions intended to benefit Black Americans, other racial minorities (e.g., Hispanics), undefined ethnic minorities, or other underserved groups (i.e., women, gender minorities, and disabled persons). We used actions targeting Black Americans and undefined ethnic minorities to construct our dependent variables, and actions targeting other minority groups to construct our control variables.
One of us subsequently reviewed all data collected by the RAs and removed repeated and irrelevant actions. Examples of irrelevant actions included (a) reports on the firm’s diversity or CSR ratings, both of which are consequences of past actions rather than new actions; (b) announced actions intended to fulfill the terms of a conciliation agreement or resolve a lawsuit; and (c) announced appointments of directors and executives who were members of minority groups. Such appointments typically took a long time (e.g., three months) to develop and were unlikely to be a consequence of protest violence in the previous four weeks, which was the window during which we measured corporate diversity actions in response to protest violence. Table 1 presents examples of actions included in the final dataset, such as disclosing diversity data, donating to nonprofits or historically Black colleges or universities, and advocating for racial justice and equality legislation.
Examples of Corporate Diversity Action Announcements
Our final dataset includes 358 corporate diversity action announcements from 148 firms that were intended to benefit Black Americans or undefined racial minorities during our observation period. Figure 3 plots the aggregated frequency of these actions over time and the number of BLM protests over time, and Table A2 in the Online Appendix provides the number of violent and peaceful protests, corporate diversity actions, and endorsements in 2020 in our sample. The sharp increase in announcements of corporate diversity actions following the death of George Floyd on May 25, 2020 provides initial evidence that firms used corporate diversity actions to respond to the changing sociopolitical environment created by BLM protests.
Number of Protests and Corporate Diversity Actions Each Week in 2020*
Dependent Variables
Our first dependent variable, Corporate diversity actions, is a dummy variable that indicates whether a firm announced any diversity action intended to benefit Black Americans and/or undefined racial minorities in the focal week. In some press releases and media articles, companies used related terms, such as “racial diversity” and “people of color,” but did not explicitly specify “African-American” or “Black American” in their action announcements. We counted these announcements as actions because they were likely to be responses to the BLM movement given their timing and general racial and ethnic focus. We used a dummy variable rather than the count of new actions because sometimes an action can be disassembled into multiple sub-actions, which a count variable would overweight. 4 We did not use the magnitude of actions because many actions are non-financial (e.g., increasing the number of Black employees) or non-quantifiable (e.g., efforts to build a more equitable organizational culture). However, as shown in Table A5, results remain similar when we included only announcements with explicit financial commitments of more than $1 million.
Our second dependent variable, Movement endorsement, is a dummy variable that equals 1 if the firm publicly endorsed the BLM movement in its announcement of corporate diversity actions and 0 otherwise. As with the first dependent variable, we collected this information from both press releases and media articles. Endorsements took the form of publicly supporting the BLM movement or explicitly linking the firm’s actions with the death of George Floyd, the event that triggered the 2020 BLM movement. We considered an announcement to be an endorsement when (1) the firm explicitly expressed “Black lives matter” in its statement (e.g., “Black lives matter. This shouldn’t be a controversial or divisive statement. It is a universal truth.”); (2) the announcement opposed movements counter to BLM (e.g., “We came to understand that the way some, but not all, people are using the phrase ‘All Lives Matter’ in the current environment intentionally minimized the focus on the painful reality of racial inequity”); or (3) the announcement suggested that the firm’s actions were prompted by the death of George Floyd (e.g., “The killing of George Floyd has compelled our nation to once again grapple with the long history of injustice endured by Black people in America”). While we treat all three criteria as forms of endorsement, our results are similar when we used only the first criterion, as shown in Table A5.
An example of a corporate diversity action announcement that did not count as an endorsement was AbbVie’s announcement of a donation to “help address racial inequity issues plaguing our nation,” without any mention of the BLM movement. Table 1 provides additional examples of the various ways that organizations endorsed the BLM movement when announcing actions in response to the movement’s protests in their communities.
Independent Variables
Reported violence in BLM protests
We computed Reported violence in BLM protests as the number of BLM protests that were reported to have included violence per one million population in the CBSA where a firm is headquartered. We used this per capita measure so that our independent variable was scaled similarly to our moderators. Our results are consistent when we used the natural log of the number of incidents, as presented in Table A3. We selected a four-week moving window to capture this variable in order to allow businesses enough time to have debated and developed a response and to have announced it. We excluded violence in protest events with three or fewer participants because such small-scale events are unlikely to capture business leaders’ attention and do not reflect collective actions in the community (Pan & Zhang, 2019). 5
We defined violence as reports of any property damage or injuries to police officers caused by the protests. This definition aligns closely with prior research on protest violence (Soule & Earl, 2005). For example, Gamson (1990, p. 74) defined protest violence as “deliberate physical injury to property or persons,” and Wasow (2020, p. 645) defined it as “a detectable level of injury or property damage.” We excluded injuries to BLM protesters because the underlying reasons for such injuries are more ambiguous (e.g., they could be caused by policing practices) and are thus harder for outside observers to interpret (McAdam & Su, 2002). Similarly, we excluded police arrests of BLM protesters, as arrests are likely outcomes of protest violence (Davenport et al., 2011). 6
Historical protest violence
We operationalized a community’s Historical protest violence, calculated as the number of incidents of non-BLM protests with reported violence that occurred in a CBSA between January 2017 and April 2020, divided by CBSA population (per one million). We used the per capita number of historical protest violence incidents rather than the absolute number, to enable comparisons across communities of different sizes. The CBSAs with the highest levels of Historical protest violence in our sample period were Portland-Vancouver-Hillsboro (OR-WA) (8.5), St. Louis (MO-IL) (2.1), Springfield (MA) (1.6), and New Haven-Milford (CT) (1.2). As presented in Table A5, the results are similar when we used the raw number of historical non-BLM protest violence.
Historical grievance-validating events
A central grievance of the BLM movement is police brutality and racism toward Black Americans. Police shootings of Black Americans not only are recognized as a salient and deadly form of police brutality but also reflect the wider experiences of suffering and injustice faced by Black residents in their communities. Although wealth inequality could also spur grievances within the local community, it tends to attract less attention from community observers and is less relevant to the central grievance of the BLM movement (i.e., deadly police brutality). Therefore, we calculated Historical grievance-validating events as the number of police shootings of Black Americans per one million Black residents in a CBSA, minus the number of police shootings per one million for other racial or ethnic groups within a CBSA. This time-invariant variable captures all police shooting events from January 2015 (when The Washington Post began collecting these data) to April 2020 (prior to the murder of George Floyd). The CBSAs that had the most historical grievance-validating events in our sample are Calhoun (GA) (452.1), Oshkosh-Neenah (WI) (277.4), Elkhart-Goshen (IN) (177.1), and Fayetteville-Springdale-Rogers (AR-MO) (164.2). As presented in Table A5, our results remain consistent when we used the per capita number of police shootings of Black residents without adjusting for the police shootings of other racial groups.
Control Variables
We included control variables at the firm and CBSA levels that might influence firms’ Corporate diversity actions and Movement endorsement. At the firm level, we controlled for Firm size, calculated as the natural logarithm of total assets, because larger firms could face more scrutiny and be more motivated to react to changing social sentiments. We also controlled for Firm ROA (return on assets), calculated as net income divided by total assets, because higher-performing firms might have more resources available to invest in their diversity-related reputations. Additionally, some firms might have a general tendency to enact new corporate diversity practices due to the presence of social management devices (McDonnell et al., 2015). Therefore, we controlled for each firm’s Prior diversity actions related to Black Americans and Prior diversity actions related to non-Black American minorities, measured as the natural log of the number of announced actions that were intended to benefit Black Americans and other minority groups, respectively, from January 2017 through April 2020. As CEO political ideology may shape how firms react to social activism (Briscoe et al., 2014), we controlled for CEO liberalism, using contributions of each firm’s CEO to the Democratic Party minus their contributions to the Republican Party, divided by the total contribution amount (Christensen et al., 2015). Finally, we controlled for Initial diversity actions, a dummy variable that indicates whether a firm had ever taken any corporate diversity action that benefited Black Americans or undefined racial minorities between May 25, 2020 and the focal week, as firms that had already taken such actions may have been less motivated to do more.
We controlled for relevant time-variant variables at the CBSA level and national level. As local COVID-19 cases might have influenced both BLM protests and firms’ actions during our observation window, we controlled for COVID-19 cases, defined as the number of new COVID-19 cases in the previous four weeks per one million CBSA population. We also controlled for other forms of protest, namely the Number of peaceful BLM protests, the Number of counter-BLM protests, and the Number of other protests. As we did for Reported violence in BLM protests, we measured each of these variables by capturing the per capita number of protests occurring between week t–4 and t–1 in a CBSA. We defined peaceful protests as those in which no property damage or police injuries were reported, and counter-protests as those having a claim against that of BLM (e.g., “All Lives Matter” or “Back the Blue”).
As the Scale of BLM protests might lead to both protest violence and firm actions, we controlled for it using the per capita number of protesters in all BLM protests in the preceding four weeks. The CCC dataset provides multiple estimates of protest size from different sources, and we used the mean estimate calculated by CCC. Following previous research (Zhang et al., 2023), we replaced missing values with zeros, as most protests without size information were small events covered by a single source. We also controlled for Local media attention to BLM protests, using the per capita number of news articles regarding BLM protests from local news agencies in the prior four weeks in the focal CBSA. Similarly, we controlled for Local media sentiment to BLM protests, representing average news sentiments from those articles, using a fine-tuned BERT-based multilingual sentiment classification model (Devlin et al., 2019). Including these variables helps to control for the alternative explanation that the association between protest violence and firm responses is mediated by media attention and sentiment to violence. To ensure that our results are not driven by Violent policing of BLM protests (Ives & Lewis, 2020), we controlled for the per capita number of protests in the preceding four weeks in which police deployed chemical agents, i.e., tear gas and other chemical irritants.
We controlled for the Percent of local firms taking action, defined as the percentage of local firms that announced at least one corporate diversity action in the prior four weeks, to account for possible diffusion of responses across firms in a community (Davis & Greve, 1997). Moreover, we controlled for the Number of accumulated local BLM protests with violence and the Number of national BLM protests with violence, measured as the per capita number of BLM protests with violence in the focal CBSA from May 25, 2020 to week t–5, and the total number of BLM protests with violence at the national level outside of the focal CBSA. We included a control for the Percent of industry peer firms taking action, which is the proportion of industry peer firms that took corporate diversity actions after May 25, 2020. This variable helps us to address industry-level contagion in corporate practices as an alternative explanation (Marquis & Tilcsik, 2016).
In addition, we controlled for several time-invariant variables at the CBSA level. This included the Event community, a dummy variable that equals 1 for firms headquartered in Minneapolis-St. Paul-Bloomington (MN), the CBSA where George Floyd was murdered, and 0 otherwise. As shown in Table A5, our results are consistent when we removed firms located in Minneapolis-St. Paul-Bloomington. As communities with certain characteristics may be more likely to observe BLM protests and protest violence, we controlled for Community population (by one million) and the percentages of Community Black residents, Community less-than-college residents (i.e., less than a bachelor’s degree), Community young residents (under age 30), and Community rural residents. We also controlled for Community liberalism (Forti et al., 2025), measured as the percentage of voters in the CBSA who voted for the Democratic presidential candidate, Hillary Clinton, in the 2016 presidential election. To account for the economic disparity that Black Americans may encounter in a community, we controlled for the Poverty rate differences between Black residents and all other residents in a community, reported in the American Community Survey. Finally, we controlled for the Number of firms in the community, specifically the raw number of firms in the same CBSA as the focal firm, because communities with more firms may be viewed as more favorable places for organizing protests.
Estimation Technique
To test our hypotheses, we used a linear probability model (LPM) with panel data. We employed a panel design rather than a cross-sectional design because variation in both Reported violence in BLM protests and Corporate diversity actions occurs both across CBSAs and within CBSAs over time. Aggregating firm–week observations to a single observation per CBSA would prevent us from modeling this temporal variation within CBSAs. Additionally, the panel structure allowed us to establish temporal precedence by lagging the independent variable and including time-varying controls, thereby mitigating concerns about reverse causality.
Because industry differences and national-level trends may influence a firm’s tendency to take action in response to BLM protests, we included industry (based on two-digit SIC code) and week fixed effects in our regressions. We did not include CBSA fixed effects because doing so would preclude us from testing the main effects of all time-invariant variables, including our moderators. 7 We prefer LPM over probit and logit models because recent research (Timoneda, 2021) has shown that LPM with fixed effects yields more-accurate estimates and predicted probabilities than maximum likelihood estimation does when the dependent variable frequency is less than 25 percent; Corporate diversity actions, our first dependent variable, is a rare event that represents 2.5 percent of firm–weeks in our sample. Employing week and industry fixed effects in probit or logit models would eliminate 18.8 percent of observations when predicting Corporate diversity actions and 40.8 percent of observations when predicting Movement endorsement, which could lead to estimation bias (Berk, 1983). However, our results are largely consistent when we used the logit model with sector (based on one-digit SIC code) and week fixed effects, as presented in Table A5.
To test Hypotheses 2 and 4, we examined whether firms announced corporate diversity actions in the wake of protests and whether they publicly endorsed the movement in those announcements. Hence, only firms that made announcements in a given week were included in the sample when we tested these two hypotheses. Unobserved factors influencing the inclusion of firms in this sample could cause sample selection bias in the models used for these tests. To address this issue, we adopted the Heckman two-stage model to correct for sample selection bias (Heckman, 1979), similar to approaches used in related research (Fremeth et al., 2022; Wowak et al., 2022). In our setting, this approach uses a first-stage probit regression to predict the likelihood of firms announcing Corporate diversity actions and calculates an inverse Mills ratio that is controlled for in a second-stage regression.
The selection equation of the Heckman two-stage model requires a valid exclusion restriction, a variable that affects sample selection but not the outcome (Certo et al., 2016; Wolfolds & Siegel, 2019). In our case, this means identifying variables that affect whether a firm announced corporate diversity actions but not whether it endorsed the movement. We used two such variables: Firm ROA and Prior diversity actions related to non-Black American minorities, both of which are defined as control variables in the previous section.
The rationale for ROA is that firms with stronger financial performance face less immediate pressure to deliver short-term results and may therefore feel more freedom to invest in stakeholder-oriented actions (DesJardine & Bansal, 2019; Flammer & Bansal, 2017). However, once a firm has decided to take such actions, the additional cost of publicly endorsing BLM is relatively low, making it unlikely that ROA directly influences endorsement decisions. Similarly, firms with a history of diversity-related actions targeting other minority groups are likely to have more experience and infrastructure (e.g., relationships with relevant NGOs) to support new initiatives. Yet, after controlling for a firm’s orientation toward supporting Black Americans, these capabilities should not directly affect the decision to explicitly associate new actions with the BLM movement.
We also assessed the empirical validity of these two exclusion restrictions. As expected, the coefficients on the exclusion restrictions, Firm ROA and Prior diversity actions related to non-Black American minorities, are statistically significant (p < 0.001 for both) in predicting Corporate diversity actions, indicating their relevance in predicting selection. When we used both variables to predict Movement endorsement, their coefficients were not significant (β = −0.735, p = 0.149 and β = 0.031, p = 0.590, respectively), suggesting that they are both exogenous variables. Based on these results, we calculated the inverse Mills ratio (IMR) from the first-stage probit model and controlled for it in the second-stage model predicting Movement endorsement.
Results
Main Results
Table 2 presents the descriptive statistics and correlations for all variables in our study. Models predicting Corporate diversity actions and Movement endorsement based on the Reported violence in BLM protests in a firm’s local community are presented in Tables 3 and 4 (Table 4 shown further below).
Descriptive Statistics and Correlations*
The correlation between Corporate diversity actions and Movement endorsement is missing because the latter has a value only when the former equals 1.
Preliminary Regressions Predicting Corporate Diversity Actions and Movement Endorsement*
p < .05; ••p < .01; •••p < .001
Each observation represents a firm–week. Standard errors are clustered at the CBSA level.
Regression Results for Hypothesized Moderating Effects on Corporate Diversity Actions and Movement Endorsement*
p < .05; ••p < .01; •••p < .001
Each observation represents a firm-week. Standard errors are clustered at the CBSA level.
Table 3 summarizes a collection of preliminary regression results. Model 1 presents the regression results of the first-stage probit regression predicting Corporate diversity actions for the Heckman two-stage model. Models 2 and 3 include Reported violence in BLM protests, our focal independent variable in predicting Corporate diversity actions. Model 3 shows that the main effect of Reported violence in BLM protests is positive and non-significant (β = 0.022, p = 0.150) after including control variables. Looking at the significant control variables, we see that corporate diversity actions tended to be announced by larger firms, firms with stronger financial performance, firms that had taken diversity actions before May 2020, and firms in communities where a higher percentage of local peers took actions in the preceding four weeks.
We find it interesting that Local media attention to BLM protests is negatively correlated with Corporate diversity actions, which could imply that business leaders were hesitant to take stances on contentious social issues if they inferred that such actions would be profiled by the local media. We also find a positive, albeit insignificant, effect of CEO liberalism, which points in the direction of prior research suggesting that liberal CEOs tend to respond positively to social activism causes (Briscoe et al., 2014; Chin et al., 2013). Additionally, while Table 2 shows that the Scale of BLM protests is positively correlated with Reported violence in BLM protests, Corporate diversity actions, and Movement endorsement, the multivariate models show that the Scale of BLM protests is not a significant predictor of Corporate diversity actions or Movement endorsement. Looking beyond political ideologies, Chung et al. (2025) suggested that business leaders hesitate to take positions on social issues that are risky or contentious, as was the case with BLM. This underscores the need to account for the community context of the event, which likely differs from less-contentious issues and events such as the Women’s March movement (Zhang et al., 2023).
Model 4 reports the results without controls when predicting Corporate diversity actions, providing initial support for our Hypotheses 1 and 3. Model 5 reports results without controls when predicting Movement endorsement, indicating initial support for Hypothesis 2 but not Hypothesis 4.
Table 4 reports regression results that formally test our hypotheses. Hypothesis 1 proposes that a community’s history of protest violence weakens the positive effect of reported violence on the likelihood of Corporate diversity actions. Model 6 shows that the interaction term Reported violence in BLM protests × Historical protest violence has a significant negative effect on the likelihood of Corporate diversity actions (β = −0.006, p = 0.000), supporting our prediction. Hypothesis 2 suggests that Historical protest violence in a community moderates the impact of Reported violence in BLM protests on firms’Movement endorsement. This hypothesis is supported by the results of Model 7, which indicate that the interaction between Reported violence in BLM protests and Historical protest violence has a significant negative effect on the probability of firms’Movement endorsement (β = −0.031, p = 0.000). 8
Hypothesis 3 posits that a community’s history of grievances related to a protest issue strengthens the positive effect of reported violence on local firms’ responses to the protest. We tested this by interacting Reported violence in BLM protests with Historical grievance-validating events. As shown in Model 8, Reported violence in BLM protests has a stronger effect on Corporate diversity actions in communities with more Historical grievance-validating events (β = 0.0005, p = 0.043), supporting Hypothesis 3. Hypothesis 4 posits that Historical grievance-validating events in a community strengthen the effect of Reported violence in BLM protests on firms’Movement endorsement. Inconsistent with our expectation, Model 9 shows that the effect is not significant (β = −0.001, p = 0.476).
Model 10 includes both interaction terms when predicting Corporate diversity actions, and the results continue to support both Hypothesis 1 (β = −0.008, p = 0.000) and Hypothesis 3 (β = 0.001, p = 0.000). As presented in Model 11, when both interaction terms are included, the results continue to support Hypothesis 2 (β = −0.031, p = 0.000) but not Hypothesis 4 (β = −0.000, p = 0.937) when predicting Movement endorsement.
To visualize these effects, we plotted the results from Model 10 in Figures 4 and 6 and the results from Model 7 in Figure 5. As shown in Figure 4, the marginal effect of Reported violence in BLM protests on Corporate diversity actions is weaker for companies headquartered in communities with a high level of Historical protest violence (0 + 2 SD, or 1.06) than for companies headquartered in communities with no Historical protest violence. 9 In a community with no record of Historical protest violence, each unit of Reported violence in BLM protests increases the probability of Corporate diversity actions by 1.27 percentage points. At the global average probability of Corporate diversity actions of 2.47 percentage points (mean value across all CBSAs and times in our sample), this marginal effect corresponds to a 51.4 percent increase relative to the baseline probability (1.27/2.47). In contrast, in a community with high Historical protest violence (0 + 2 SD), that same unit increase in Reported violence in BLM protests translates to a 0.43 percentage point increase in the probability of Corporate diversity actions, corresponding to a 17.4 percent increase relative to the baseline probability (0.43/2.47). Figure 5 plots the marginal effect of Reported violence in BLM protests on Movement endorsement. As shown, this effect is larger in communities with no Historical protest violence than in communities with a high level of Historical protest violence (0 + 2 SD, or 2.23), decreasing from 11.03 to 4.09 percentage points. At the global average probability of Movement endorsement of 21.79 percentage points, the marginal effect corresponds to a 50.6 percent increase (11.03/21.79) for communities with no Historical protest violence and an 18.8 percent increase (4.09/21.79) for communities with a high level of Historical protest violence.
Conditional Marginal Effect of Reported Violence in BLM Protests on Corporate Diversity Actions by Historical Protest Violence (H1)
Conditional Marginal Effect of Reported Violence in BLM Protests on Movement Endorsement by Historical Protest Violence (H2)
Conditional Marginal Effect of Reported Violence in BLM Protests on Corporate Diversity Actions by Historical Grievance-Validating Events (H3)
As illustrated in Figure 6, the effect of Reported violence in BLM protests on Corporate diversity actions is highly positive for companies headquartered in communities with a high level of Historical grievance-validating events (mean + 1 SD, or 61.21), whereas it turns negative in communities with a low level of Historical grievance-validating events (mean −1 SD, or −4.03). For communities with a high level of Historical grievance-validating events (mean + 1 SD), the marginal effect of Reported violence in BLM protests is 3.26 percentage points. At the global average probability of Corporate diversity actions of 2.47 percentage points, this marginal effect corresponds to a 132.0 percent increase (3.26/2.47). In contrast, in a community with a low level of Historical grievance-validating events (mean −1 SD), that same unit increase in Reported violence in BLM protests decreases the probability of Corporate diversity actions by 1.13 percentage points, corresponding to a 45.7 percent decrease (–1.13/2.47).
As shown in Table 5, we predicted the marginal effect of Reported violence in BLM protests for eight CBSAs in the U.S., based on regression results from Model 10 (for Hypotheses 1 and 3) and Model 11 (for Hypotheses 2 and 4). For instance, in San Jose-Sunnyvale-Santa Clara (CA), where the total population is 1.94 million, a one-unit increase in the incidence of protest violence per million residents translates to 1.94 more such events in the past four weeks. According to our calculations, this increase raises the predicted probability of Corporate diversity actions by 3.20 percentage points, from a global average of 2.47 percentage points to 5.67 percentage points. Similarly, it raises the predicted probability of Movement endorsement by 11.02 percentage points, increasing from a global average of 21.79 percentage points to 32.81 percentage points. These figures demonstrate that the effect of protest violence is economically significant in a community such as the San Jose-Sunnyvale-Santa Clara CBSA, which has a low level of Historical protest violence and a high level of Historical grievance-validating events. In contrast, in Washington-Arlington-Alexandria (DC-VA-MD-WV), where Historical protest violence is high and Historical grievance-validating events is low, a one-unit increase in Reported violence in BLM protests reduces the likelihood of Corporate diversity actions by 0.17 percentage points and increases the likelihood of Movement endorsement by only 9.26 percentage points.
Predicted Changes in Corporate Diversity Actions and Endorsements Following an Increase of One Violent BLM Protest per Million Population, by CBSA*
High and low values for Historical protest violence and Historical grievance-validating events are defined as above or below the sample median, respectively. Predictions are calculated using regression results from Model 10 and Model 11.
Robustness Analyses
We conducted several robustness tests in addition to those referenced above. Results from the following tests are reported in Table A3 in the Online Appendix. First, we used the log-transformed number of incidents of Reported violence in BLM protests rather than the per capita number to capture our independent variable. As shown in Models 1–2 of Table A3, Hypothesis 1 (β = −0.014, p = 0.001), Hypothesis 2 (β = −0.058, p = 0.000), and Hypothesis 3 (β = 0.001, p = 0.001) are supported when the alternative measure is used, while Hypothesis 4 (β = 0.003, p = 0.123) is not supported at the 0.1 level.
Second, as our independent variable is at the CBSA level, we collapsed the unit of analysis to the community–week instead of the firm–year. To conduct this community-level analysis, we mimicked our main analysis and averaged all the firm-level control variables within the same community. As a result, our sample size decreased from 14,508 (for Hypotheses 1 and 3) and 358 (for Hypotheses 2 and 4) to 2,914 and 249, respectively. Shown in Models 3 and 4 of Table A3, Hypothesis 1 (β = −0.022, p = 0.000), Hypothesis 2 (β = −0.029, p = 0.000), and Hypothesis 3 (β = 0.002, p = 0.000) remain supported when we used community–week as the unit of analysis, while Hypothesis 4 (β = 0.001, p = 0.173) is not supported.
The Effects of Peaceful Protests
Although we focus on violence in protests, we also examined whether peaceful protests produce a similar pattern of results and whether the effect of protest violence is significantly stronger than that of peaceful protests. To do so, we first interacted the variable Number of peaceful BLM protests (from our control variables) with both moderators, Historical protest violence and Historical grievance-validating events, and added these interaction terms to the models that include the interactions with Reported violence in BLM protests, so we could directly compare the effects of violent and peaceful protests. We did not include Hypothesis 4 here as it is not supported in our main analysis. As shown in Models 5–7 of Table A3, Hypothesis 1 (β = −0.005, p = 0.001), Hypothesis 2 (β = 0.000, p = 0.050), and Hypothesis 3 (β = −0.023, p = 0.022) remain supported when these terms are included, supporting our argument that reported protest violence requires a community-based interpretation.
To further test whether violent protests require more community-based interpretation than peaceful protests do, we split our sample based on the median value of Historical protest violence and Historical grievance-validating events. Reported in Table A4 in the Online Appendix, results from the split sample analysis are consistent with Hypotheses 1 and 2, showing that the effect of Reported violence in BLM protests becomes weaker and even reverses in the CBSAs where Historical protest violence was high (i.e., above the sample median). We also find that Reported violence in BLM protests has a positive effect in the CBSAs where Historical grievance-validating events was high and a negative effect in CBSAs where the value was low (i.e., below the sample median), which is directionally consistent with Hypothesis 3.
We also compared the effects of violent and peaceful protests in these split sample regressions. In CBSAs with low Historical protest violence, violence has a larger positive coefficient than the Number of peaceful BLM protests does for Corporate diversity actions (difference = 0.037, p = 0.001) and for Movement endorsement (difference = 0.423, p = 0.001). However, in CBSAs with high Historical protest violence, the relative effect of violence becomes smaller and statistically indistinguishable from that of peaceful protests for Corporate diversity actions (difference = −0.050, p = 0.133) and is significantly more negative than that of peaceful protests for Movement endorsement (difference = −0.520, p = 0.000). Regarding the sample split on Historical grievance-validating events, coefficient differences between violent and peaceful protests are directionally consistent with Hypothesis 3. The coefficient difference is negative when Historical grievance-validating events is low (difference = −0.011, p = 0.209) and positive when Historical grievance-validating events is high (difference = 0.022, p = 0.178), although both differences are statistically insignificant at the 0.05 level. Overall, these findings imply that violent protests require more community-based interpretations than peaceful protests do to trigger corporate responsiveness.
Discussion
Our article offers a new framework to help scholars better understand how businesses interpret and respond to violence in their environments. When violence is reported in community protests, leaders of locally headquartered businesses may feel compelled to respond, which places them in a position requiring that they make sense of the violence before taking corporate action. Unlike other informational attributes of large-scale protests (e.g., Negro & Olzak, 2019; Wouters & Walgrave, 2017; Zhang et al., 2023), violence has uniquely polarizing significance, as it can be interpreted in one of two fundamentally opposing ways: as a reflection of social disorder, which discourages corporate action, or as a manifestation of valid community grievances, which engenders a more sympathetic response. These opposing interpretations, combined with the visibility and urgency of violence in the streets, drive business leaders to consider additional contextual information to justify their stance and, ultimately, to shape their company’s response. Our framework positions community context as a critical lens through which business leaders view and interpret reported violence in local protests.
We applied our contingent framework to announcements of corporate diversity actions in the wake of the 2020 Black Lives Matter movement. Since this movement’s central grievance was police violence against Black citizens, observers did not necessarily consider the occurrence of violence in certain local protests to be surprising or novel (Davenport et al., 2011; Ives & Lewis, 2020). And yet, importantly, interpretations of that violence sharply diverged, casting it either as evidence of community social disorder or as evidence of valid grievances in the community.
We found empirical support that community characteristics shaped the effect of reported protest violence for companies headquartered in those locations. Firms were less likely to announce diversity actions in response to protest violence in communities with significant histories of violence in protests on other issues. This finding is consistent with the notion that business leaders perceive current protest violence as a continuation of the community’s previous social disorder rather than as an expression of a strongly felt new grievance, therefore making the violence appear less justified. Conversely, firms were more likely to announce diversity actions in response to reported violence in BLM protests in communities with more historical police shootings of Black citizens, incidents that validate the grievances expressed via the current protest violence and make the violence appear more warranted.
These community characteristics also shaped business leaders’ decision either to endorse the BLM movement when announcing a supportive diversity action or to issue a supportive action without directly affiliating with the movement. While prior research has demonstrated that affiliating with activists invites further activism (McDonnell et al., 2015; Odziemkowska & McDonnell, 2024), our framework highlights the unique risks associated with endorsing a protest movement that has been visibly associated with violence. Such endorsement may provoke disapproval from broader stakeholders and public audiences, who may interpret the violence differently or interpret a corporation’s response as being supportive of violence. Among firms that announced supportive actions, local community characteristics that sustained the social disorder interpretation reduced firms’ likelihood of endorsing the movement, while community characteristics that sustained the valid grievance interpretation increased the odds of endorsement. By demonstrating how the local environment influences corporate rhetoric and alignment with activists, our framework offers new insights into the strategic calculations underlying corporate responses to contentious social movements.
A New Vantage on Corporate Responses to Violence in Communities
Many companies face rising violence in workplaces and in the communities where they operate (Mobasseri, 2019), due to factors that are yet to be fully understood but include post-pandemic shifts in behavior, a looming mental health crisis, rising levels of societal inequality, and political violence (Lombardi et al., 2024; Ortiz et al., 2022). Violence erupts, often unpredictably, at multiple levels—among individuals, within organizations and communities, and even across entire nations. Reflecting this broader trend, reports of violence in community protest movements are also on the rise in recent years, accounting for 17.8 percent of total protests globally in 2006 and 21.9 percent in 2020 (Ortiz et al., 2022). Yet, despite the growing prevalence of violence, how companies interpret and respond to it remains poorly understood. A common assumption is that business leaders will instinctively distance their companies from acts of violence, given that violence is often viewed as illegitimate and unlawful by society and stakeholders. However, corporate decision makers have learned to be attuned to shifting moral sentiments in the communities where they operate (e.g., Yue et al., 2013). This sociopolitical sensitivity raises the possibility that business leaders could view community protest violence in different ways, perhaps believing that in certain circumstances it is partly justified and deserving of a corporate response or even corporate endorsement. Examining how business leaders make this determination, our framework puts community context front and center.
Our community-embedded theory of how business leaders respond to violence has implications beyond community protests. For example, how an organization such as a hospital struggling with workplace violence (Doehring et al., 2024; Wang et al., 2021) chooses to respond may depend on how its leaders understand their community’s history with respect to sources of violence and those grievances to which violence in the community is attributed. This “community” may be the hospital itself or the broader geographic region in which the hospital operates. Multinational corporations navigating geopolitical crises, such as armed conflicts or civil unrest, must decide whether to withdraw, engage in de-risking strategies, or take an active stance aligned with community sentiment (Saittakari et al., 2023). Moreover, as protests on the wars in Gaza and Ukraine have revealed, companies of different sizes and types are pressured to take positions on geopolitical issues following violence in protests. Finally, digital platforms confronting online violence, such as harassment or extremist rhetoric, must determine whether to suppress or tolerate certain forms of expression based on their perception of the legitimacy of those forms of expression in the community. By accounting for community context, our framework explains why businesses may not always instinctively distance themselves from violence but may, instead, engage in nuanced, morally and strategically driven responses.
Contributions to Research on Social Movements and Markets
Our theory and findings also advance research on how social movements shape markets (Durand & Georgallis, 2018; Hiatt et al., 2015; McDonnell et al., 2015; Yue et al., 2013). First, we offer a new way to think about the contingent nature of organizational leaders reacting to reports of community protests, which contributes to the literature on the activism-as-information logic (King & Soule, 2007; Zhang et al., 2023). While some protest features, like size or frequency, may appear to be relatively straightforward clues regarding a social issue’s relevance to the community (Amenta et al., 2010; Wouters & Walgrave, 2017), violence is inherently ambiguous. For a protest movement that includes violence, the implication is that size or frequency alone is insufficient to predict how organizations will respond: Larger and more frequent community protests may prompt supportive organizational responses (Ferguson et al., 2018; Negro & Olzak, 2019; Zhang et al., 2023) if the protests are peaceful, but if violence is reported, supportive responses may hinge on how local history can inform the justifications for that violence. The same nuance can apply to other characteristics of social movements that have largely shown a one-sided linear effect on stakeholder responses, such as social movement organization membership, the legitimacy of activists, and issue breadth (Fassiotto & Soule, 2017; Hiatt et al., 2009; Rao et al., 2003).
Second, by applying our violence-interpretation framework to the business decision to endorse or not endorse the BLM movement, we provide a theoretical bridge connecting protest features with business endorsements of social movement organizations. While many scholars have investigated how movements influence organizational policies and practices to align with movement goals (Ingram et al., 2010; King, 2008; Soule et al., 2014), few have examined factors that lead businesses to associate themselves publicly with a movement, for instance, by explicitly joining a movement’s boycott campaign (McDonnell, 2016) or publicly collaborating with activists (Odziemkowska, 2022; Odziemkowska & McDonnell, 2024). Extending previous work on the reputational benefits of selectively affiliating with movements, we show how local histories influence these decisions. Although prior studies have suggested that business leaders have reasons to avoid being seen as cooperating with activists tied to disruption (Briscoe et al., 2015; Rojas, 2006) or violence (Orazani et al., 2021; Thomas & Louis, 2014), our study reveals conditions that make leaders more likely to take this risk and endorse a reportedly violent movement.
Third, our theorizing about interpretations of reported violence can also inform broader social movement research on responses to protest violence, which includes conflicting findings that violence can increase (Enos et al., 2019; Gamson, 1990) or decrease (Bailey et al., 2023; Rojas, 2006; Thomas & Louis, 2014) the effectiveness of protests in achieving movement goals. By emphasizing the role of the local community in which reported protest violence occurs, we have shown how other community characteristics and events provide valuable contextual information about issues that can shape interpretations of protest violence, ultimately affecting observers’ responses to the movement. An executive of a major U.S. corporate retailer shared with us the following thoughts about such contextual information:
Our response at the beginning of Black Lives Matter was . . . we’re going to evaluate what matters most to our stakeholders. We’re going to look at the communities that we serve . . . We’re going to see where we think we have gaps, and then we’re going to make a plan on how to respond to the issue. It was a very stressful time . . . You feel this pressure, and I think that’s what happens with local events that really get a lot of ground swell.
We suspect that, as reflected in this executive’s thinking, the contextual backdrop of local community characteristics is particularly important for protests associated with violence, as they trigger greater scrutiny from business leaders and community members. Indeed, results from our supplemental analyses suggest that reports of protest violence require more context-based interpretation than reports of peaceful protests do.
Finally, our results linking protest violence across different protested issues in the same community offer a new direction for research on social movement spillovers (Meyer & Whittier, 1994; Wang & Soule, 2012). We found that a record of reported violence in past community protests can weaken the effectiveness of subsequent protests on other social issues. This suggests that when violence persists across multiple and different historical protest movements supporting different social issues in the community, it can start to appear endemic, such that business leaders and local stakeholders may conclude that violence is no longer a meaningful indicator of grievances. For social movement mobilization research, this suggests that in places where chronic grievances and threats fuel repeated violent mobilization (Almeida, 2019; Chae et al., 2024; LeFebvre & Armstrong, 2018; McKane & McCammon, 2018), the ensuing protest pattern may create a barrier to achieving movement outcomes in the form of limiting corporate response. Our study implies that past protest violence can have a negative long-term effect on the effectiveness of other movements, adding to a small but important body of research on indirect community-based spillover effects across movements (see Ferguson et al., 2018).
Contributions to Research on Stakeholder Strategy and Corporate Activism
Our study also contributes to existing research by documenting street protests as a driver of corporate activism. In recent years, firms have increasingly taken stances on contentious social issues, and employees, customers, and other stakeholders often respond when they do so (Bondi et al., 2025; Qin et al., 2025; Yue et al., 2024). An executive we interviewed explained that “there’s been a heightened intersection of politics and business decision-making in the last five years . . . which I’ve seen change very dramatically.” While researchers have theorized about how legal, organizational, stakeholder, and leader attributes may influence such corporate sociopolitical activism (Chung et al., 2025; Hambrick & Wowak, 2021; McKean & King, 2024; Wowak & Busenbark, 2024), little is known about how external civil society influences a company’s decision to engage in activism. Our arguments suggest that local protests can play a role, prompting corporate announcements of organizational actions that explicitly endorse social movements and their contentious causes. Indeed, as shown in Figure 3, in response to BLM protests, an unprecedented proportion of companies announced new corporate diversity actions, some of which were explicitly linked to the BLM movement. By showing that violence in street protests can lead to corporate activism and under what conditions, our study sheds light on the antecedents of corporate activism.
Directions for Future Research
Several of our study’s limitations provide fruitful avenues for future research. While we analyzed firms’ immediate reactions to reported protest violence, it is worth investigating whether and how these reactions lead to concrete long-term benefits for Black Americans. Future research could explore the extent to which corporate diversity actions in the wake of reported protest violence concretely advance the movement’s racial equity goals (Baker et al., 2024; Kim & Lyon, 2015; Marquis et al., 2016).
Our understanding of the business response to violence can also be extended in several ways. Researchers could consider how organizational structures and roles, such as community engagement officers, shape the interpretation of violence in local protests, workplaces, and communities broadly. In addition, while we studied the largest U.S. corporations, scholars could theorize how smaller local organizations respond to community protest violence (Knight et al., 2022; Negro & Olzak, 2019), as they tend to be more embedded in and dependent on their local communities for survival. A company’s own direct history with community protesters may also alter its interpretations of reported violence in community protests. For example, firms that have themselves been targeted by violent activists in the past may use a different calculus when deciding how to respond to reported violence in street protests, being more alert to the risks of violence targeting their own operations in the future. Finally, by studying other protest movements, scholars could further explore the heterogeneity of corporate actions taken in response to reports of protest violence, including reactionary responses such as withdrawal from communities and markets or even support for alternative or counter movements.
On this front, we do not interpret our findings as evidence that violence generally elicits supportive corporate responses, beyond our research context of the 2020 BLM movement. A substantial literature concludes that violent tactics often undermine a movement’s credibility and reduce support or concessions—including in civil conflict settings—because they signal disorder and undermine movement legitimacy (e.g., Bailey et al., 2023; Rojas, 2006; Thomas & Louis, 2014). Moreover, we do not claim a universal positive effect of reported violence; instead, effects emerged only where local histories provided cues that legitimized the grievance, and were muted where histories made violence seem endemic. We therefore refrain from claiming generalizability to other violent movements; rather, we suggest that future research center on community context when evaluating whether and how reported violence shifts organizational responses, which will help to establish the boundary conditions of the effects of violence.
Finally, because corporations are embedded in both local and national contexts, it is also important to understand how firms respond to local versus national protests and how these responses relate to firms’ local versus national profiles. In a more global context, certain national institutional conditions may also be needed for our theory to apply. For example, if the local press lacks freedom, business leaders may doubt that media reports on violence contain useful information, or they may learn to rely on other sources to decipher community sentiment regarding contentious social issues.
Supplemental Material
sj-pdf-1-asq-10.1177_00018392261419416 – Supplemental material for Interpreting Violence: How Community Context Shapes Corporate Responses to Street Protests*
Supplemental material, sj-pdf-1-asq-10.1177_00018392261419416 for Interpreting Violence: How Community Context Shapes Corporate Responses to Street Protests* by Forrest Briscoe, Mark R. DesJardine and Muhan Zhang in Administrative Science Quarterly
Footnotes
Acknowledgements
We thank Associate Editor Christina Ahmadjian and three anonymous reviewers for helping us improve this paper. We also thank Donald Hambrick, Sarah Soule, Martin Kilduff, Brayden King, John McCarthy, and members of the Slump Management reading group for helpful feedback and discussions. Earlier versions of this paper were presented at Pennsylvania State University, Peking University, Chinese University of Hong Kong, Hong Kong Baptist University, Cornell University, the University of Michigan, the Future of Activism workshop at Northwestern University, the Alliance for Research on Corporate Sustainability conference in Washington, D.C., the Academy of Management conference in Boston, and the European Group on Organization Studies conference in Cagliari. We gratefully acknowledge funding support from the Smeal College of Business, the Smeal Center for the Business of Sustainability, and the Penn State Africana Research Center.
Data Availability Statement
The data used in this study was compiled from public and proprietary sources. Licensing restrictions prevent us from publicly posting the resulting datasets. The manuscript identifies all data sources, and we included a detailed data collection process for information not provided by commercial data sources. A detailed variable-to-source mapping and regression code is available from the authors on request.
Supplementary Material
1
For example, consider two communities: one with a history of peaceful previous-movement protests and one with a history of violent previous-movement protests. In the first community, the effect of protest violence in the current movement would be large, greatly increasing firms’ likelihood of new supportive corporate actions. In the second community, by contrast, the effect of protest violence in the current movement would be lesser, only slightly increasing firms’ likelihood of new supportive corporate actions or even potentially reducing their likelihood.
4
For instance, a firm might announce that it will invest $100 million to address racial diversity issues or that it will donate $50 million to the National Association for the Advancement of Colored People (NAACP) and $50 million to historically Black colleges and universities. While the impact of the $100 million donation might provide similar benefits to Black Americans, the number of actions would be twice as large for the second announcement if we were to use a count (and even larger if the company further subdivided its donations).
5
6
Including the per capita number of protests involving protester injuries and protester arrests as additional control variables does not change our findings.
7
8
9
For low levels of Historical protest violence, we took the value of zero rather than two standard deviations below the mean since the latter would produce nonsensical negative values.
Authors’ Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
