Grievance-based social movement mobilization in the #Ferguson Twitter storm

Abstract

Existing literature on collective action suggests that social protest activity is often driven by structural out-group grievances. This article explores how a framework of grievance-based social movement participation applies to the digital media realm and how social media are reshaping the protest landscape. Our research looks specifically at the case of the #Ferguson Twitter storm that occurred in November 2014. During a 3-week period, over 6 million tweets were sent with the indicator #Ferguson. We examine the statistics and content of those tweets to show that the Ferguson Twitter storm was driven to an enormous volume by four key mobilizers. Tweet content included structural out-group grievances that reflect established expectations about drivers of social movements and protests. In contrast to the emphasis on violence by traditional mass media, online social movement participants emphasized peace, especially after the conflict escalated and rioting in the streets began.

Keywords

Grievances protest social media social movement Twitter

The month of November 2014 was tense in Ferguson, Missouri. Citizens were waiting for the grand jury to decide whether or not to indict a White police officer, Darren Wilson, for shooting and killing an unarmed Black teenager, Michael Brown. Some witnesses said the teenager attacked the police officer and tried to take his gun. Other witnesses said Officer Wilson was the aggressor. The grand jury had to sort through conflicting witness accounts, changing stories, and little physical evidence to determine whether police officer Wilson should be charged with a crime.

On 24 November at 8:30 pm, St Louis County prosecuting attorney Robert McCulloch announced that the grand jury had decided not to indict Officer Wilson. This triggered a chain reaction of protests across the nation. Protesters blocked bridges and tunnels, crowded out onto roadways, and disrupted post-Thanksgiving shopping in more than 150 cities across the country. Protests in Ferguson turned violent with the destruction of police cars and local businesses.

During the 3 weeks surrounding the announcement on 24 November, social media activity increased dramatically with the hashtag indicator #Ferguson appearing in over 6 million tweets on Twitter. What messages were prevalent in #Ferguson tweets? What conclusions can we draw from data about message resonance and user participation? Does dialog associated with #Ferguson reflect traditional conflict drivers cited by social movement scholars? This article starts to address some of these questions by examining the details and trends in those 6 million tweets and comparing the online dialog regarding the Ferguson events to established theories of conflict, social movements, and protests.

Literature review

Literature on social movements, protest, and collective action is vast. Van Stekelenburg and Klandermans (2013: 92) note that “At the heart of every protest are grievances.” We know, however, that not all aggrieved groups protest (Orum, 1974). In addition to the presence of grievances, issues of identity, opportunity, resources, emotion, and existing networks all play a role in the initiation and escalation of protest events. With the advent and expanding access to social media platforms and the development of digital networks, scholars have begun to examine the role of new technologies in facilitating protests.

Protest motivators and accelerators

Protest participation may be motivated or accelerated based on a number of factors such as grievances, identity, resources, efficacy, and individual traits. We examine each of these factors in further detail below.

Grievances

Grievances, as main drivers for protest participation, are generally understood as the perception that a group or person is experiencing illegitimate inequality, has been the victim of injustice, or feels a general moral indignation about some state of affairs (Klandermans, 1997). These feelings cannot always be resolved through official channels. When a poor fit exists between a group’s interests and the capacity of established political institutions to address those interests, social movement action and protests become more likely (McVeigh, 1995). Experiments and observations from actual protests have shown that when citizens feel other groups have a relatively unjust advantage, their tendency to see protest as an appropriate response increases (Leach et al., 2006; Van Zomeren et al., 2004). Along with this, procedural justice and the level of fairness in relative and relational aspects of the social process are strong predictors of protest participation (Morrison, 1971; Tyler and Smith, 1995). When grievances become compounded, the predictive power of groups to protest increases (Kriesberg and Dayton, 2012). Likewise, individuals experiencing low rank in multiple status areas such as education and income often perceive a cumulative effect of those lower statuses (Bradburn and Caplovitz, 1965). Finally, when grievances are directly attributed to a structural out-group such as political leadership, law enforcement, or other institutional organization, protest behavior becomes increasingly attractive, as more normative modes of engagement are perceived as lacking efficacy, and institutional attributions become increasingly pronounced (Simon and Klandermans, 2001; Smyth, 2002).

Identity

Many factors in addition to grievances contribute to protest participation. Identity, and social identity in particular, is important because it is often tied to grievances and reinforces distinctions that exacerbate feelings of relative deprivation (Van Zomeren et al., 2004). When these identities become politicized, they often include grievances and connect feelings of inequality and unfairness with a causal, external enemy (Van Stekelenburg and Klandermans, 2013). Further, when such identities are salient, the ability to empathize and cooperate with out-groups makes traditional channels of negotiation and cooperation less plausible (Smyth, 2002). “The more people identify with a group, the more likely they are to protest on behalf of that group” (Van Stekelenburg and Klandermans, 2013: 890).

Resources

Along with identity, access to resources is a strong indicator of protest likelihood (Tilly, 1978). The types of resources available will also determine both the type of action and the choice of location for protest events (Klandermans, 1984). When a group effectively utilizes the resources at their disposal, collective action efforts occur more rapidly and effectively (Tarrow and Tollefson, 1994; Tilly, 1978). These resources can include money, supporters, and attention (Davis et al., 2005). Resources also include the presence and depth of interpersonal networks (Schussman and Soule, 2005). Networks also assist in dissemination of information about protest events and provide locations where aggrieved citizens can collaborate and share information critical of existing structures (Paxton, 2002; Van Stekelenburg and Klandermans, 2013).

Efficacy

Efficacy, or an individual’s expectation that it is possible to alter conditions through protest (Van Stekelenburg and Klandermans, 2013), also informs participation. This phenomenon has several layers. Group efficacy is the perception that collective protest efforts will influence the outcome of a problem (Bandura, 1997). Political efficacy outlines general faith in the political process as well as individual understanding of the political circumstances (Van Stekelenburg and Klandermans, 2013). Finally, individual efficacy will influence protest participation, particularly when individuals feel that their unique skills are critical to the success of a protest event (Klandermans, 1984). For example, civic skills, (Schussman and Soule, 2005), which are unique to individuals, can serve as critical resources, especially for those involved in leadership and strategy development.

Individual traits

Several other factors have been shown to increase the likelihood of protest participation. Factors unique to individuals such as levels of extroversion and internal locus of control (Margetts et al., 2015) have been connected to a willingness to engage in social movements. Basic biographical characteristics are also important (Schussman and Soule, 2005). Lack of family and related responsibilities, disposable income, and flexible employment circumstances all minimize potential limitations to protest participation.

Mainstream media and alternatives

Mainstream media coverage of American social movements, protest events, and civil disobedience has been studied by academics in great detail since the 1960s when battles for civil liberties raged and activism was pronounced. Scholarship on the subject notes mass media tendency to emphasize violence and the sensational facets of protests (McCurdy, 2012). Rucht (2004) suggests that social movements often attempt to counter this type of message with their own narrative and emphasis on opposing ideals. Similarly, Boyle and Schmierbach (2009) found that mainstream media were consistently critical of protest activity and that comparatively, alternative media were positively associated with protest participation and online activism. Some research, however, still shows that citizens are more likely to trust information from mainstream media sources than social media messages (Ceron, 2015).

The role of social media in protest and social movements

The role of social media platforms in the evolution of protest processes has been both lauded and minimized. The Arab Spring provided academics with a wealth of information on how citizens utilize new technologies to advance protest movements. While initial enthusiasm and causal claims have waned, many scholars now acknowledge that these new connective platforms must be factored in when discussing how citizens express discontent with existing power structures and citizen–state relations (MacKinnon, 2012). They also note that platforms like Facebook and Twitter “have certainly opened up innovative avenues for people to challenge existing configurations of power” (Biekart and Fowler, 2013: 529).

Connectivity and social networking are credited with three main areas of influence in the protest process: capturing public attention, the evasion of censorship, and logistics and coordination of protest events (Tufekci, 2014: 3). Because state-sponsored and mass media are no longer the gatekeepers of information, public attention is more easily captured through citizen-driven information dispersion. Likewise, the speed of information dispersion within social networks allows protesters to adjust quickly to government action, police movement, safety threats, and other challenges (Gerbaudo, 2012).

Though not enough research yet exists, some also believe that these new tools facilitate more democratic, inclusive forms of dialog (Biekart and Fowler, 2013). Because citizens who had previously been denied a voice now have an outlet to express themselves, situations can be understood with more context and less reliance on elites for information. Beckwith (2012) also proposes that we are approaching a global citizenry that shifts away from the existing “me” mindset and toward a “we” mindset because of new access. By removing geographic barriers, we become less self-focused and more interested in the well-being of those outside our immediate reach.

Not everyone is as optimistic about social media’s contribution to the protest process, however. Bennett and Segerberg (2012) note that while recent movements that relied on social media platforms for collaboration have been successful in deposing or disrupting regimes, lasting peace and long-term change remain elusive. Tufekci (2014: 12) notes,

By allowing protesters to scale up quickly, without years of preparation, digital infrastructure acts as a scaffold to movements that mask other weaknesses, especially collective capacities in organizing, decision-making, and general work dynamics that only come through sustained periods of working together.

Though protests before the digital age were laborious and slow to affect change, the time allowed for effective coordination, consensus building, and planning have been notably absent in more recent protests. Finally, in the developing world, the resource-poor still cannot participate online to the same degree, or at all (Norris, 2002; Scheufele and Nisbet, 2002).

Digital participant characteristics

Because social media platforms allow for speedy information dissemination and low cost forms of communication, they have been utilized by many groups hoping to mobilize citizens for physical engagement in protest events. Additionally, social media platforms serve as an outlet to those unable or uninterested in participating in physical protests allowing them to remain engaged with the subject of those protests and their broader social implications.

What is the relation between online mobilization and offline protest? In a study on Belgian protesters, Van Laer (2010) found that those working on a protest via online forums were more likely to have higher levels of education and greater interest in politics than physical protest participants. He also found that those without existing strong ties to official organizational networks offline were unlikely to be mobilized to participate in protests based on online content. This aligns with early claims by social media-for-activism critic Evgeny Morozov (2009) that online engagement is merely “slacktivism” and lacks effective social or political impact. As Van Laer’s (2010) study discovered, online information channels can link to offline protest to a point, but strong emotions and feelings of injustice form a route to physical protest independent of online activity.

Vissers and Stolle (2014), in a study of Canadian undergraduates, found the opposite, and that Facebook in particular appeared to be valuable in mobilizing for causes and reinforcing their importance even amongst those not previously engaged at high levels. Similarly, Bacallao-Pino (2014) found that mobilizers for the Occupy Wall Street movement believed that social media were valuable for making dissent visible. Other protesters expressed that online communication was critical, particularly hashtags that act as “powerful tools for conveying a conversation around a strategically chosen subject” (Bacallao-Pino, 2014: 5). The subjects of this study, however, expressed concern about future movements that relied predominately on social media for mobilization (Bacallao-Pino, 2014).

Another important area of new social movement research relates to the lack of hierarchy in online social movement promotion and mobilization. While traditional social movements typically offered examples of charismatic leaders and well-defined hierarchies (Oberschall, 1973; Zald and Ash, 1966), new movements are less likely to follow this pattern. In fact, some movements vocally and actively reject traditional conceptions of leadership (Sutherland et al., 2013).

Theory

Clearly, we are yet to reach a stage where we can confidently draw conclusions about social media platforms, protests, and users working within the online sphere to advance physical protest events and associated social movement efforts. Early reliance on anecdotal stories about the impact of social media rightly solicited criticism and failed to accurately represent the diverse nature of online/offline interactions during protest events. As we begin to build a body of knowledge about the role of social media in protests, we must capitalize on the enormous quantity of digital information these platforms provide.

Our research proposes a case study on the large body of Twitter information generated during the Ferguson protests. We theorize that much of the established literature on social movement mobilization and protest triggers will be observed within the digital social movement dialog. In particular, we expect to see evidence of grievances against structural out-groups (law enforcement and the justice system), the salience of social identity (race relations), and the impact of resources (online social networks in the form of Twitter followers). At the same time, not all participants in high profile social media movements are actively engaged with the topic. The public attention brought to certain issues online can attract neutral news outlets or general internet loafers who join in without true interest in the topic. The key question is whether social movement leaders can create a central voice and amplify their message above the white noise of ad hoc online activity.

We believe that traditional conceptions of “designated leaders” (Fairhurst and Grant, 2010) will be less visible. Instead, we expect to observe newer leadership trends as defined by Sutherland et al. (2013: 11) where “a leadership actor may be understood as any individual who exercises power by managing meaning, defining reality and providing a basis for organizational action.” In this sense, we expect that resources, such as large audiences, may create inadvertent leaders, and that designated leaders and those actively engaged with the movement may not always enjoy the same degree of message resonance as inadvertent leaders whose audience size serves as an important digital resource. Additional research by Liu et al., (2014) suggests that original content on Twitter is increasingly being replaced by retweets. The timing of the protest events in Ferguson align with this observed increase in retweets and reinforces our expectation that inadvertent leadership will result from increased retweeting of content from those with larger audiences and more relative resources.

Finally, we posit that digital engagement with the hashtag indicator #Ferguson may indicate tension between citizen perceptions of events in Ferguson and mass media coverage of those events—in particular, riots and violence. Previous research suggests that mass media will emphasize violence (Boyle and Schmierbach, 2009) and that one tactic social movement leaders take to counter such tendencies include creation of an alternative message (Rucht, 2004). We believe that a push for peace will be evident as an alternative message especially in the time-period immediately following the grand jury announcement and extensive mass media coverage of rioting in Ferguson.

In order to uncover these theoretical characteristics of the online digital protest, we divide our exploration into three main areas:

Identify mobilizers and leading voices. In line with Liu et al.’s (2014) observations about increased retweet behavior, we expect that the Ferguson Twitter storm was driven to high volume by a small set of key mobilizers with substantial digital resources (Twitter followers). We expect to find that leaders within this movement may not fall within traditional definitions of leadership, and instead will align more with Sutherland et al. (2013) and their emphasis on leadership that is iterative with a focus on leaders as those who define meaning in a movement.

Identify motivators and accelerators. Based on established literature of social movement mobilization and grievance, we expect to find that structural out-group grievances against police and the justice system are prevalent in tweet content and that social identity issues emerge in discussions of race relations.

Identify alternative messages. Building on Rucht’s (2004) alternative message theory, we hypothesize that tweet activity escalated in terms of promoting peace over violence in the few days after the grand jury announcement in contrast to emphasis on riots and violence by mainstream media outlets.

Methods and data

Methodology selection for exploring such a vast amount of Twitter data was critical for our project. We wanted to collect a complete set of data for quantitative analysis, and we also needed a complete analysis of the content and tone of tweets to uncover our hypothesized trends. In this section, we describe the method used for collecting tweets and associated metadata and the method for categorizing content and tone of the tweets.

A complete set of Twitter data was collected for 3 weeks from 14 November 2014 to 4 December 2014. We wrote a program in Python to access the Twitter streaming application interface (API) and ran the program around the clock to capture tweets in real time and save them to a database. All tweets that contained #Ferguson were saved along with metadata associated with each tweet including screen name of the user, number of followers, and whether the tweet was a retweet. In total, we collected more than 6 million tweets for this timeframe. Figure 1 charts the number of tweets per day based on Greenwich Mean Time (GMT) day boundaries showing a clear spike in activity after the announcement of the grand jury decision at 8:30 pm local time on 24 November.

Figure 1.

Millions of tweets per day for #Ferguson, divided into phases.

Quantitative analysis was performed on the complete dataset of 6 million tweets to create a set of descriptive statistics useful in analyzing overall trends. We created scatter plots of all unique users in the dataset based on factors including number of tweets and number of retweets. We also summarized the dataset by type of tweet and sorted the dataset to bubble up the top tweets and tweeters in various categories.

A random sampling of the tweets was then coded for content themes using independent coders. The tweet timeline was divided into three phases as shown in Figure 1: the period prior to the grand jury announcement (pre-storm), the 3-day “Twitter storm” period right after the grand jury announcement (storm), and the period after that (post-storm). A random sampling of 1000 tweets was taken from each of the phases. The 3000 sampled tweets were coded with two independent coders per tweet in order to facilitate content analysis, sentiment analysis, and comparison across samples. We created a codebook of the most frequently occurring themes and recruited 18 undergraduate students to code the tweets per the codebook. There were nine themes divided into three categories: (1) the mention of structural out-group grievances (police conduct, race relations, and social justice); (2) the overall sentiment (supportive, critical, or neutral toward the particular issue); and (3) calls for others to take action (violent, peaceful, or digital action).

Sentiment mining, or opinion analysis, has emerged as a field of its own and focuses on the computational treatment of sentiment and opinion as derived from text (Pang and Lee, 2008). Sentiment mining, or opinion analysis, can be difficult when interpreting a microblog limited to only 140 characters. For example, one tweet said “RT @stevegiegerich: Protesters briefly shut down So Grand/44. Police keep distance. #Ferguson http://t.co/m2ixJpmhX0,” which included a photo of protestors on the highway. This tweet was interpreted as critical of police conduct by one coder and supportive of police conduct by another. Was it good or bad that the police kept their distance? A number of studies have been done to develop models for automatically detecting mood through text analysis (Bollen et al., 2011). Often these techniques require preprocessing of the data subjectively and elimination of Uniform Resource Locators (URLs) that contain links to photos, videos, or other websites. For purposes of this study, we wanted to include all components of each tweet including URL links and photos, so we used human coders. This generated some disagreement which we resolved as missing data and considered content themes present only when both coders agreed.

Findings

In this section, we present our findings for the three areas of exploration as defined in the Theory section above.

Identifying mobilizers and leading voices

Although close to 1.5 million unique users participated in the #Ferguson Twitter storm, most participants (61%) tweeted only once during the 3-week period. Less than 1% of the users tweeted more than ten times per day. Figure 2 demonstrates these extremes of user participation. One user tweeted almost 14,000 times, whereas almost a million users tweeted only once. The drop-off in participation occurs rapidly from the high-volume tweeters to a long tail of users tweeting less than 100 times total.

Figure 2.

User participation by number of tweets.

Twitter messages can be divided into two broad categories: an original message created by the user and a retweet that passes another user’s message along. A retweet often represents agreement or support for the original message. In the 3-week period we include for analysis, the vast majority of tweets (78%) were retweets, and only 22% were original tweets. The daily percentage of original tweets was consistent across the phases ranging from 19–25%. People were most likely to participate in this Twitter storm by simply retweeting another user’s message.

The idea of “top tweeters” can be conceptualized in different ways, and we will explore three of them. First, we consider “top tweeters” as highly active users who tweeted the most original messages. These people generate original content, rather than merely echoing another user’s sentiment. Appendix 1 shows the top 20 original tweeters. Some of the users on this list, including the most active tweeter, have very few followers, and thus lack the ability to reach a broad audience.

We then consider “top tweeters” by retweet rate. In doing so, we also account for audience receptiveness to messages and the reach of the original tweeter. This notion of influence aligns with Berger and Strathearn’s (2013) study of White supremacist activists and their definition of influence as the “tendency of a user to inspire measurable reaction from others” (p. 4). Appendix 2 shows the top 20 most retweeted users, which includes mainstream news organizations such as Cable News Network (CNN), British Broadcasting Corporation (BBC), and Fox News, which have 14 million, 12 million, and 5 million followers, respectively. The list contains basketball celebrity Kobe Bryant, with almost 6 million followers. The other top tweeters have fewer but still substantial followings and tend to be other news agencies, self-described protesters, and special accounts set up specifically for Ferguson.

Finally, we conceptualize “top tweeters” by combining the two views discussed above. Figure 3 shows a scatter plot of users by their original tweets versus how often they were retweeted. Although most of the users fall into the lower left corner, some areas of the graph stand out. Bubbling up toward the top are four users we call “key mobilizers.” These users (deray, Bipartisanism, YourAnonNews, and The AnonMessage) actively generated new messages and their messages were retweeted in very large volumes. A second category, which we call “unwitting mobilizers,” tweeted only once or twice, but their messages were heavily retweeted. Kobe Bryant tweeted only once, and his tweet was retweeted over 19,000 times. This demonstrates the potential power for celebrities to impact online social media dialog. The third category, which we call “moderate mobilizers,” includes users who were actively tweeting and actively being retweeted at a level that made them stand out from the crowd in the lower left of the graph. Finally, the fourth category stuck out in the lower right. These users which we call “passionate participants” tweeted more original tweets than any other user, yet their messages were rarely retweeted.

Figure 3.

Scatter plot of most influential users.

Identifying motivators and accelerators

Results from the coding of our random sample of 3000 tweets were evaluated to determine intercoder reliability as shown in Table 1. The intercoder observed agreement values show that coders were in agreement more than 82% of the time when coding for calls to action, but had low levels of agreement in the other coding areas. Cohen’s kappa for each pair of students ranged from 0.1 to 0.4 indicating only slight to fair agreement (Landis and Koch, 1977).

Table 1.

Intercoder reliability for coding of tweet samples.

Content theme	Observer agreement	Cohen’s kappa
Topic: police conduct mentioned	74%	.37
Topic: race relations mentioned	66%	.21
Topic: justice mentioned	62%	.13
Tone: negative/positive about police conduct	62%	.32
Tone: negative/positive about race relations	60%	.20
Tone: negative/positive about justice	60%	.27
Call to action: violent	92%	.15
Call to action: peaceful	86%	.25
Call to action: digital	82%	.10

The coded content reflected a significant presence of structural out-group grievances and social identity issues. Almost 40% of tweets mentioned grievances around police conduct, race relations, and/or the justice system as shown in Table 2. The tone of the tweets was overwhelmingly coded as negative (critical) for each of the three themes with four times as many negative as positive or neutral tweets. Some of the tweets expressed outrage at the systematic problems around police conduct, such as “Down with the Opressing Police and Police brutality!!! #Ferguson.” The negative tone for the theme of race relations strongly voiced structural problems that still exist for African Americans. An example tweet is: “White privilege is real. White supremacy is deadly. Black liberation is anti-white supremacy not anti-white. #Ferguson.” Tweets about justice within our legal system were also highly negative, such as: “We need to turn our judicial system upside down. Murder can’t be legalized because of our racist societal illness. #Ferguson.” While these tweets all clearly express a grievance, those grievances are firmly grounded in adversarial attributions tied to identity. This supports Simon and Klandermans’ (2001) observation that identity-based grievances are often blamed on a specific out-group and that an authority or “the system” is responsible for the suffering of in-group members.

Table 2.

Frequency of structural out-group grievance themes.

Theme	% of total tweets	% of coder agreed tweets
Police conduct	17% (497/3000)	22% (497/2212)
Race relations	14% (423/3000)	21% (423/1973)
Justice system	13% (399/3000)	22% (399/1844)
At least one of above	36% (1085/3000)	39% (1085/2803)

The three themes of structural out-group grievances appeared regularly across the 3-week period under study including all three phases of the Twitter storm as shown in Figure 4. The topics appeared in 10–20% of the tweets in each phase, although the prevalence of the themes of police conduct and race relations dropped off during the 3-day storm right after the grand jury announcement. The percentage of justice-related tweets remained the same during that time-period when there were 10 times more tweets in total reflecting the focus on the justice system at that point.

Figure 4.

Prevalence of structural out-group grievance topics by phase.

If we look beyond the random sample and consider all 6 million tweets, we can explore the content by reading the most retweeted tweets. Appendix 3 shows the top five most retweeted messages across the 3-week timeframe. All five messages bring out the systematic complaint of police use of force targeted against young, Black men. However, the most retweeted message is one of peace and reconciliation showing a young, Black protestor giving “free hugs” to a White police officer. That leads us to our next set of findings around tweets encouraging peace over violence.

Identifying alternative messages

We hypothesized that tweets promoting peaceful action would be more likely to occur in the days following the grand jury announcement, in contrast to the escalation of violence that was seen in the streets of Ferguson and emphasized in mass media coverage during the same timeframe. Tables 3 and 4 show the results of cross-tabulation analysis of coding of calls for violence and peace, respectively, in random samples for each phase of the Twitter storm. Interpreting across the “Promoted” row in Table 3 for violent action, we see that 2.2% of tweets in the pre-storm phase promoted violence compared to 0.3% in the storm phase. The likelihood of promoting violent action was about seven times lower after the grand jury announcement. Interpreting across the “Promoted” row in Table 4 for peaceful action, we see that 3.2% of tweets in the pre-storm phase promoted peace compared to 5.5% in the storm phase. The likelihood of promoting peaceful action was one and a half times higher after the grand jury announcement. Even though the number of tweets in the coded sample is small, we can see that our hypothesis is directionally supported.

Table 3.

Phase dependence of tweets promoting violence.

Violent action	Pre-storm	Storm	Post-storm	Total
Promoted	19	3	4	26
	(2.2%)	(0.3%)	(0.4%)	(0.9%)
Not promoted	850	921	973	2744
	(97.8%)	(99.7%)	(99.6%)	(99.1%)
Total	869	924	977	2770
	(100.0%)	(100.0%)	(100.0%)	(100.0%)

X²: 21.2; df: 2; p: .001.

Table 4.

Phase dependence of tweets promoting peace.

Peaceful action	Pre-storm	Storm	Post-storm	Total
Promoted	28	45	28	101
	(3.2%)	(5.5%)	(3.2%)	(3.9%)
Not promoted	859	778	845	2482
	(96.8%)	(94.5%)	(96.8%)	(96.1%)
Total	887	823	873	2583
	(100.0%)	(100.0%)	(100.0%)	(100.0%)

X²: 7.8; df: 2; p.02.

In addition to promotion of peaceful action, the tweets in the “storm” phase also escalated in calls for digital action, a form of non-violent action. Digital action was coded whenever a tweet was promoting online action such as retweeting, liking, or posting on other social media. Table 5 shows the cross-tab analysis of the call for digital action in tweets, which was seven times more likely to occur during the days immediately following the grand jury announcement than before.

Table 5.

Phase dependence of tweets promoting digital action.

Digital action	Pre-storm	Storm	Post-storm	Total
Promoted	6	33	14	53
	(.7%)	(4.9%)	(1.5%)	(2.2%)
Not promoted	844	638	923	2405
	(99.3%)	(95.1%)	(98.5%)	(97.8%)
Total	850	671	937	2458
	(100.0%)	(100.0%)	(100.0%)	(100.0%)

X²: 34.7; df: 2; p: .001.

Discussion

This detailed look at a large Twitter storm provides insight into the nature of participants and the structure of resulting collective action. There are several elements that come together to form an online social movement, including structural out-group grievances, a triggering event, and key mobilizers. These elements can be viewed as part of a timeline of several stages of a social protest which include a preparation phase, an ignition phase, and a protest phase (Hussain and Howard, 2013). We will discuss each of these three phases individually as they relate to the #Ferguson activity.

In the preparation phase, activists form their online social networks and establish their goals. This is an essential stage in the online protest movement, as activists already need a well-established following prior to the ignition phase. A number of key individuals and organizations existed prior to the Ferguson event and had a large following on Twitter. Most of the top 20 retweeted users during the Ferguson Twitter storm as shown in Appendix 2 had large Twitter audiences, and many had associated organizations and established goals. Some were large news agencies such as CNN, Fox News, and the BBC. Others were self-described protesters and activists seeking social justice such as “deray,” Talbert Swan, and Anonymous. Some activists formed particularly to address the Michael Brown case in Ferguson and built their following during the pre-storm phase and the months prior, such as the user “mikebrowncover.” Our findings showed that structural out-group grievances of police conduct, race relations, and social justice for African Americans were prevalent throughout the preparation or pre-storm phase of the protest. This aligns with Smyth’s (2002) discussion of social identity salience. Because structural out-group grievances were becoming increasingly salient, cooperation through official channels within existing institutions was becoming less plausible.

The role of activists and their associated organizations is fairly straightforward during the preparation phase as they work to build their message, network, and following. The role of news agencies and celebrities during this phase is less so, as they have goals and interests that are broader than any single social issue. Online social movements may turn to news agencies or celebrities to help bring attention to issues, but such attention comes with some risk. News agencies may follow more objective journalistic guidelines telling both sides of a story and potentially diluting the message of a cause. Celebrities come with their own agenda and concern for their public persona which could drown out the core message with a variant that does not align with the goal of the cause (Tufekci, 2014). Activists may proceed with caution in their use of journalists and celebrities as focusers of attention, but they ultimately have no control if such individuals choose to enter the dialog of their own accord. In the case of the #Ferguson Twitter storm, contributions of celebrities such as Kobe Bryant were on-point and helpful to activists for racial justice.

In the ignition phase, a symbolic yet powerful event occurs that galvanizes the public around the structural out-group grievances. In this case, that event was the grand jury announcement of no indictment for the police officer who shot Michael Brown. This event was not only a grievance itself for many but was also symbolic of overarching structural out-group grievances in society. This trigger point was a clear marker in the Twitter space for #Ferguson. The quantity of #Ferguson tweets spiked to ten times previous levels and the nature of tweets was distinctive during the few days immediately following the event. The prevalence of some structural out-group grievances (police conduct and race relations) declined during the ignition phase, or “storm” phase as we call it, but the prevalence of social justice dialog continued at the same level. Tweets that promoted violent action dropped off during the ignition phase while tweets promoting peaceful action increased, both with statistical significance. The focus of tweets narrowed in scope and increased in intensity, and that focus was predominantly about the need for justice and the need to pursue justice peacefully.

In the protest phase, protesters strategically organize for social or political change. In the protest, or post-storm phase, prevalence of all three structural out-group grievance themes returned to the same levels as the preparation phase. Promotion of violent action, however, did not return to the previous level, but stayed as low as the ignition phase. Promotion of digital action decreased from the ignition phase, but was twice as high as during the preparation phase. The on-going online protest turned away from violent action and toward digital action even as riots were occurring in the streets of Ferguson. This aligns well with Van Laer’s (2010) study indicating strong feelings of injustice can drive a different route for physical protest than online protest. Both the ignition phase and protest phase were influenced by the key mobilizers identified in the Twitter storm. Those four key mobilizers were actively promoting their messages and these were retweeted in high volumes.

One other important discussion point involves the methods that were used in this study for content analysis and sentiment analysis. We felt, and social media scholars agree, that it was essential to consider photos, videos, and other links provided in the tweet URLs when doing content analysis and sentiment analysis (Berger and Strathearn, 2013). Sentiment analysis that only looks at text can miss key elements of the message. In our 3-week study, 71% of the 6 million tweets contained a URL. Our method of random sampling and use of independent human coders produced useful results in evaluating the nature of the online activity. The intercoder disagreement does prompt further research into how people interpret tweets. After all, communication is not only about the sentiment of the message sender, but also about how the message is decoded and interpreted by the receiver.

Finally, one of the limitations of our methodology comes from data collection based on a single hashtag, #Ferguson. We did not include hashtags that surfaced as the movement grew including #BlackLivesMatter, #ICantBreathe, #HandsUpDontShoot, and others. Even though it is not practical to capture all dialog of an issue across all of Twitter, analysis done on individual hashtags is important, as it provides a foundation for comparative work. Additionally, the enormous use of #Ferguson warrants its individual examination, as it was listed by Twitter as the second most used hashtag worldwide for the year 2014 (Wagstaff, 2014).

Conclusion

Our analysis of the #Ferguson Twitter storm provides a valuable snapshot of users, messages, and themes that resonated across Twitter during the timeframe prior, during, and following the grand jury results. The Ferguson Twitter storm was driven to a huge volume by four key mobilizers with high levels of involvement and digital resources along with a handful of other mobilizers with moderate influence. Participation grew to include about 1.5 million people, although the majority of those people tweeted only once. Content themes found in the tweets included structural out-group grievances against the police force and justice system and social identity issues regarding race relations. In these ways, the digital dialog reflected established expectations about likely drivers of traditional grievance-driven social movements and protests. In contrast to the emphasis on violence by traditional mass media, online social movement participants were more likely to encourage peace, especially after the conflict escalated and rioting in the streets began.

Beyond this specific case, the methods utilized here are useful as social scientists develop and refine methods for investigating digital social interactions. Our study demonstrates various modes for evaluating user influence and analyzing content and sentiment that provide a baseline case for future comparative work on other social media platforms and protest events. Our study does raise several areas where further research is warranted. First, we need a better understanding of individual efficacy in digital protests and why certain types of messages or users seem to resonate more than others online. Second, we should work to understand the role of different types of attachments within online social media messages. Do pictures, news articles, personal videos, or other attachments influence retweet frequency? Finally, what are the connections between the preparation, ignition, and protest phases of online social movements? We need to understand citizen grievances from digital social movement analysis, but we also need to understand how those stated grievances move the dialog through each of these phases and eventually shape change in state infrastructure and policy. As Van Laer (2010: 413) points out, “The question, however, whether participation in collective action mediated through digital information channels can indeed be turned into real sustained commitment, still remains open for further research.” We believe our study has taken an important step in understanding the online social movement surrounding the events in Ferguson, Missouri in 2014. The methods we employed proved useful in identifying key grievances and mobilizing messages for social change. Connecting such analysis to “real sustained commitment” will be an on-going effort.

Footnotes

Appendix

Appendix 3.

Top 5 most retweeted Tweets.

Count	Tweet text	Link expanded
20,031	RT @TheBlackGuyX: A 12 yr old protesting held a sign that read “Free Hugs”. The officer asked if he could have 1 … #Ferguson http://t.co/hqVvNltrkn
18,827	RT @kobebryant: The system enables young black men to be killed behind the mask of law #Ferguson #tippingpoint #change
17,412	RT @voice: Don’t stop spreading this until @BarackObama sees it. He must hear OUR VOICE. #ferguson http://t.co/jIQdTNyvYL
12,317	RT @Sierra2231: My mom said it best. #Ferguson http://t.co/jAu28dq6sy
11,897	RT @CrystalLewis: In case you still don’t know why there’s so much outrage in #ferguson … (via @creativerobd) #handsup #DONTSHOOT http://t.co/lNXtBsSwCT

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author biographies

Rebecca Kay LeFebvre has her PhD in International Conflict Management and is a lecturer in the Department of Political Science & International Affairs at Kennesaw State University (KSU). She has an MS in Computer Science and combines her knowledge of technology and conflict to study the impact of digital media on democratic governance and collective action.

Crystal Armstrong is a PhD candidate in International Conflict Management at KSU. Her research focuses on the formation of social identities in the digital sector and the role of digital social identities in cyber-conflict.

References

Bacallao-Pino

(2014) Social media mobilisations: articulating participatory processes or visibilizing dissent? Cyberpsychology: Journal of Psychosocial Research on Cyberspace 8(3). DOI: 10.5817/CP2014-3-3

Bandura

(1997) Self Efficacy: The Exercise of Control. New York: W.H. Freeman.

Beckwith

(2012) The birth of a global citizenry: fantasy or future reality. In: Hubbard

(ed.) Birth 2012 and beyond: Humanity’s great shift to the age of conscious evolution. USA: Shift Books, pp. 155–157.

Bennett

Segerberg

(2012) The logic of connective action: digital media and the personalization of contentious politics. Information, Communication & Society 15(5): 739–768.

Berger

Strathearn

(2013) Who matters online: measuring influence, evaluating content and countering violent extremism in online social networks, 28 March. London: International Centre for the Study of Radicalisation and Political Violence. Available at: http://icsr.info/wp-content/uploads/2013/03/ICSR_Berger-and-Strathearn.pdf

Biekart

Fowler

(2013) Transforming activisms 2010+: exploring ways and waves. Development and Change 44(3): 527–546.

Bollen

Pepe

Mao

(2011) Modeling public mood and emotion: twitter sentiment and socio-economic phenomena. In: Proceedings of the fifth international AAAI conference on weblogs and social media (vol. 11), Barcelona, 17–21 July, pp. 450–453. Menlo Park, CA: AAAI Press.

Boyle

Schmierbach

(2009) Media use and protest: the role of mainstream and alternative media use in predicting traditional and protest participation. Communication Quarterly 57(1): 1–17.

Bradburn

Caplovitz

(1965) Reports on Happiness. Chicago, IL: Aldine.

10.

Ceron

(2015) Internet, news, and political trust: the difference between social media and online media outlets. Journal of Computer-Mediated Communication 20(5): 487–503.

11.

Davis

McAdam

Scott

et al . (eds) (2005) Social Movements and Organizations. New York: Cambridge University Press.

12.

Fairhurst

Grant

(2010) The social construction of leadership: a sailing guide. Management Communication Quarterly 24(2): 171–210.

13.

Gerbaudo

(2012) Tweets and the Streets: Social Media and Contemporary Activism. London: Pluto Press.

14.

Hussain

Howard

(2013) What best explains successful protest cascades? ICTs and the fuzzy causes of the Arab Spring. International Studies Review 15(1): 48–66.

15.

Klandermans

(1984) Mobilization and participation: social-psychological expansions of resource mobilization theory. American Sociological Review 49(5): 583–600.

16.

Klandermans

(1997) The Social Psychology of Protest. Oxford: Blackwell.

17.

Kriesberg

Dayton

(2012) Constructive Conflicts: From Escalation to Resolution. Lanham, MD: Rowman & Littlefield.

18.

Landis

Koch

(1977) The measurement of observer agreement for categorical data. Biometrics 33: 159–174.

19.

Leach

Iyer

Pedersen

(2006) Anger and guilt about in-group advantage explain the willingness for political action. Personality and Social Psychology Bulletin 32(9): 1232–1245.

20.

Liu

Kliman-Silver

Mislove

(2014) The tweets they are a-changin’: evolution of twitter users and behavior. In: Proceedings of the eighth international AAAI conference on weblogs and social media (vol. 30), Ann Arbor, MI, June 2014, pp. 5–14. Palo Alto, CA: AAAI Press.

21.

McCurdy

(2012) Social movements, protest and mainstream media. Sociology Compass 6(3): 244–255.

22.

MacKinnon

(2012) The Netizen. Development 55(2): 201–204.

23.

McVeigh

(1995) Social structure, political institutions, and mobilization potential. Social Forces 74(2): 461–485.

24.

Margetts

John

Hale

et al . (2015) Leadership without leaders? Starters and followers in online collective action. Political Studies 63(2): 278–299.

25.

Morozov

(2009) Brave new world of slacktivism. Foreign Policy 19(5).

26.

Morrison

(1971) Some notes toward theory on relative deprivation, social movements, and social change. American Behavioral Scientist 14(5): 675.

27.

Norris

(2002) Democratic Phoenix: Reinventing Political Activism. Cambridge: Cambridge University Press.

28.

Oberschall

(1973) Social Conflict and Social Movements. Englewood Cliffs, NJ: Prentice Hall.

29.

Orum

(1974) On participation in political protest movements. Journal of Applied Behavioral Science 10(2): 181–207.

30.

Pang

Lee

(2008) Opinion mining and sentiment analysis. Foundations and Trends in Information Retrieval 2(1–2): 1–135.

31.

Paxton

(2002) Social capital and democracy: an interdependent relationship. American Sociological Review 67(2): 254–277.

32.

Rucht

(2004) The quadruple ‘A’: media strategies of protest movements since the 1960s. In: Van De Donk

Loader

Nixon

et al . (eds) Cyberprotest: New Media, Citizens and Social Movements. London: Routledge, pp. 29–56.

33.

Scheufele

Nisbet

(2002) Being a citizen online: new opportunities and dead ends. Harvard International Journal of Press-Politics, 7(3): 55–75.

34.

Schussman

Soule

(2005) Process and protest: accounting for individual protest participation. Social Forces 84(2): 1083–1108.

35.

Simon

Klandermans

(2001) Politicized collective identity: a social psychological analysis. American Psychologist 56(4): 319–331.

36.

Smyth

(2002) Identity-based conflicts: a systemic approach evaluation project. Negotiation Journal 18(2): 147–161.

37.

Sutherland

Land

Böhm

(2013) Anti-leaders (hip) in social movement organizations: the case of autonomous grassroots groups. Organization 21: 759–781.

38.

Tarrow

Tollefson

(1994) Power in Movement: Social Movements, Collective Action and Politics. Cambridge: Cambridge University Press.

39.

Tilly

(1978) From Mobilization to Revolution. Reading, MA: Addison-Wesley.

40.

Tufekci

(2014) Social movements and governments in the digital age: evaluating a complex landscape. Journal of International Affairs 68(1): 1–18.

41.

Tyler

Smith

(1995) Social justice and social movements. In: Gilbert

Fiske

Lindzey

(eds) Handbook of Social Psychology (vol. 4). Boston, MA: McGraw–Hill, pp. 595–629.

42.

Van Laer

(2010) Activists online and offline: the internet as an information channel for protest demonstrations. Mobilization: An International Quarterly 15(3): 347–366.

43.

Van Stekelenburg

Klandermans

(2013) The social psychology of protest. Current Sociology 61(5–6): 886–905.

44.

Van Zomeren

Spears

Fischer

et al . (2004) Put your money where your mouth is! Explaining collective action tendencies through group-based anger and group efficacy. Journal of Personality and Social Psychology 87(5): 649–664.

45.

Vissers

Stolle

(2014) Spill-over effects between Facebook and on/offline political participation? Evidence from a two-wave panel study. Journal of Information Technology & Politics 11(3): 259–275.

46.

Wagstaff

(2014) #Winners: ellen, World Cup score big on Twitter in 2014. NBC News, 10 December. Available at: http://www.today.com/money/twitters-top-tweets-retweets-2014-ellen-world-cup-score-big-1D80349123

47.

Zald

Ash

(1966) Social movement organizations: growth, decay and change. Social Forces 44(3): 327–340.