Is the Whole World Watching? PDF Building a Typology of Protest Coverage on Social Media

JOURNALISM STUDIES 2020, VOL. 21, NO. 11, 1590–1608 https://doi.org/10.1080/1461670X.2020.1776144 Is the Whole World Watching? Building a Typology of Protest Coverage on Social Media From Around the World Summer Harlow a, Danielle K. Brown b...

JOURNALISM STUDIES 2020, VOL. 21, NO. 11, 1590–1608 https://doi.org/10.1080/1461670X.2020.1776144 Is the Whole World Watching? Building a Typology of Protest Coverage on Social Media From Around the World Summer Harlow a, Danielle K. Brown b , Ramón Salaverría c and Víctor García-Perdomo d a Jack J. Valenti School of Communication, Houston; bJohn & Elizabeth Bates Cowles Professor of Journalism, Diversity and Equality, Hubbard School of Journalism and Mass Communication, University of Minnesota - Twin Cities, Minneapolis; cSchool of Communication, University of Navarra Pamplona, Spain; dUniversidad de La Sabana, Autopista Norte de Bogotá, Colombia ABSTRACT KEYWORDS Previous research suggests that mainstream media coverage Activism; content analysis; around the world follows a “protest paradigm” that demonizes international news; news protesters and marginalizes their causes. Given the recent increase audiences; protest paradigm; social media in global protest activity and the growing importance of social media for activism, this paper content analyzes 1,438 protest- related English and Spanish news stories from around the world that were shared on social media, examining framing, sourcing, and marginalizing devices across media outlet type, region, language, and social media platform in order to create a typology of how the protest paradigm operates in an international and social media context. Results showed type of protest, location of protest, and type of media outlet were significantly related to whether news stories adhered to the protest paradigm. Social media shares were predicted by region of media outlet, English- language media, and type of protest. Over the last decade, there has been a global increase in large-scale protests around the world. The Global Database of Events, Language, and Tone (GDELT 2016) found that on- going protest frequency is higher in the 2010s than it was in the 1990s and 2000s. Given the international spike in protest activity (Carothers and Youngs 2015), it is impor- tant to understand the role of the media, particularly social media, in spreading news about protests. According to the protest paradigm, mainstream media coverage tends to demonize pro- testers and delegitimize protests (Chan and Lee 1984; McLeod and Hertog 1999), which can influence whether society will accept or reject protesters’ claims (McLeod and Hertog 1992). Considering the complex and often antagonistic relationship between main- stream media and protesters (Gitlin 1980), exploring how different types of media treat different types of protests in various countries around the world becomes all the more critical. This study of English- and Spanish-language news coverage shared on social media of global protests in 2015 aims to offer a typology for better understanding how the paradigm operates in an international and social media context. More specifically, CONTACT Summer Harlow [email protected] © 2020 Informa UK Limited, trading as Taylor & Francis Group  JOURNALISM STUDIES 1591 this content analysis examines protest-related articles published by mainstream, alterna- tive, and digital-only media outlets, and distributed via Facebook and Twitter, to test the continued relevance of the paradigm in this new media ecology, and to identify the content characteristics that might influence the extent to which social media audiences choose to engage with articles about protests. Protest Paradigm Research points to an asymmetrical relationship between the mainstream news media and protesters. Protesters need media coverage to promote their cause, but to attract media attention the protest must involve many people and employ radical tactics (Gitlin 1980; McCarthy, McPhail, and Smith 1996). At the same time, news coverage often stigmatizes protesters as deviants (McLeod and Hertog 1992, 1999). Scholars refer to this pattern of negative coverage as the protest paradigm (Chan and Lee 1984). Stories adhering to this paradigm are characterized by journalists’ reliance on official sources over voices of pro- testers, and narrative structures, such as framing, that favor conflict and the status quo (McLeod and Hertog 1999). The protest paradigm typically is attributed to the norms and routines underlying tra- ditional journalistic practices that value conflict in stories and privilege official viewpoints (McLeod 2007; McLeod and Hertog 1999). Recent research, however, questions the extent to which paradigmatic coverage is automatic, in light of the complexities of issues and identities in today’s social movements (Cottle, 2008). Some studies, especially those exam- ining the paradigm in an online context, identify factors mediating adherence to the para- digm. For example, Shahin and colleagues (2016) found that while an over-reliance on official sources and a focus on violence where characteristic of news coverage in India, China, and Brazil, use of other marginalization devices varied, suggesting that country spe- cificities diminished adherence to the paradigm outside of Western countries. Studies examining traditional media indicate that protesters advocating against the status quo are delegitimized while those protesting on behalf of the status quo are covered in more substantial ways (Gitlin 1980; McLeod and Hertog 1992). Frames and Devices Media framing can shape the public’s understanding of an event by emphasizing certain elements in a news story over others (Entman 1993). Media frames have the potential to influence attitudes and behaviors. As such, framing is an important factor in influencing whether the public will perceive a protest as legitimate (McLeod and Detenber 1999). Key components of framing include the use of story narratives (or frames), devices that portray certain characteristics of the story’s actors, and source selection. Frames Hertog and McLeod (1995, 2001) noted four main frames in media coverage of protests: riot (highlighting the conflict between protesters and society, and portraying protesters as deviants), confrontation (emphasizing the conflict between police and protesters), spec- tacle (focusing on the drama, oddity, and spectacle of the protests and protesters, includ- ing the number of protesters), and debate (emphasizing the reasons for the protest and 1592 S. HARLOW ET AL. focusing on protesters’ viewpoints and demands). The riot, confrontation, and spectacle frames usually serve to delegitimize protests and protesters (McLeod and Hertog 1992). Devices Protests can be marginalized through the use of certain devices, such as the way protes- ters’ tactics and actions are characterized (e.g., Dardis 2006). Describing protesters as violent associates lawlessness with a particular cause. However, peaceful protest descrip- tions have a counter effect. While still focusing on protesters’ actions instead of protesters’ demands, mentions of peaceful protest serve as legitimizing devices. Sourcing Mainstream news coverage relies on official sources – a product of journalistic practice – that can impact the framing of a story. The use of official sources to drive narratives also gives official sources more control over the portrayal of a story (Bennett and Segerberg 2011). Thus, relying on official sources often delegitimizes protests while inclusion of pro- testers as sources increases balance and legitimizes protesters and their claims. Protest Coverage Online Foundational studies on protest coverage examine mainstream, traditional media. However, a growing line of research that explores the paradigm in a digital context has detected some deviations in predicted coverage patterns. Harlow and Johnson’s (2011) analysis of news coverage of the 2011 Egyptian protests showed that The New York Times mostly abided by the protest paradigm, while journalists on Twitter and a citizen journalism news cite often broke free from it. In an international study comparing Twitter coverage of the Ferguson protests, Harlow (2019)) found media outlets, journalists, and the general public followed or deviated from the paradigm to varying degrees. Alternative and online-only media approach protest coverage differently. Alternative media are expected to publish stories that legitimize social movements and include perspec- tives the mainstream media ignore (Downing 2000). Similarly, online-only news outlets have been shown to cover the news differently than traditional mainstream media (Brown and Sinta 2016), suggesting a need to understand how type of media outlet might influence the way protests are portrayed in articles shared on social media. By focusing on protest cov- erage published by mainstream, alternative, and online-only media outlets that was shared via Facebook and Twitter, this present research adds to nascent scholarship by examining the paradigm in digital, international, and social media settings, in one comprehensive research project. We propose the following hypothesis: H1: Online mainstream outlets are more likely to publish protest news coverage that includes riot, confrontation, and spectacle frames than protest coverage published in online-native or alternative outlets. Protest Issue and Location Type of protest and location of protest in particular have bearing on the nature of media coverage. Mourão and Chen (2019) examined left- and right-leaning protests in 2013 and  JOURNALISM STUDIES 1593 2015, respectively, in Brazil, and found journalists were more critical of the rightist protests, contradicting previous literature that would suggest more favorable coverage. Boyle and colleagues’ (2004) found that war-related protests adhered to the paradigm more closely than protests related to social or labor issues. In their study of protest coverage across various U.S. media market types, Brown and Harlow (2019) found protest issue to be key to paradigmatic coverage in traditional media, and identified a “hierarchy of social struggle” in which protests related to racial injustice tended to follow negative coverage patterns more than other protest topics. Following these comparative approaches, the present study adds a global comparison of protest coverage to identify variance in news coverage or social media interactions according to protest type. Most of the studies examining the importance of protest type in media portrayals of protest are limited to protests in one city or one country. Still, some research suggests that country of protest can make a difference. Wittebols (1996) showed that news cover- age of protests was more favorable when protesters’ issues aligned with the govern- ment’s foreign policies. Other studies suggest that general attitudes toward protest activity can influence adherence to the paradigm, as the more protesters are seen as outsiders, the more negative coverage will be (Shahin et al. 2016; Streeck and Ken- worthy 2005). Kim and Shahin (2019) pointed out that news media ideological simi- larities could sometimes overcome national boundaries when covering certain social movements and protests. The authors argued there is an ideological parallelism among transnational media coverage that is sometimes mediated by foreign policy interests. McCluskey and colleagues (2009) found that anti-government protests in low-pluralism countries received negative coverage, and Brown et al.’s (2018) compari- son of media coverage of protests related to human rights issues in Mexico and the U.S. found that foreign protests received more legitimizing coverage in the U.S. than domestic ones. The current study pays particular attention to the importance of type and region of protest as potentially shaping adherence to the paradigm, thus furthering our understanding of the mediating factors disrupting paradigmatic coverage. With this in mind, we propose the following research question: RQ1a-d: How do protest topic and location predict use of the a) riot, b) confrontation, c) spec- tacle, and d) debate frames in protest-related articles shared on social media? Bilingual and Comparative Research Language and geographic differences can change media representation patterns. Studies have found a modulating effect of languages on news framing (Branton and Dunaway 2008; Oganian, Korn, and Heekeren 2016). Language also can have a significant effect on journalists’ selection of news topics (Lams 2016; Van Doorslaer 2009). Additionally, news coverage is affected by a country’s economic, cultural, political, and historical con- texts (Hanitzsch 2011; Weaver 1998), so geographic differences can shift representations. Framing differences have been detected when different countries cover the same news events (e.g., Machill, Beiler, and Fischer 2006; Peng 2008). Differences in news framing also stem from culturally specific journalistic practices and audience resonance (Gamson and Modigliani 1989; Van Gorp 2007). With this in mind, any typology looking to identify commonalities in protest coverage across borders must take into account linguistic and regional differences of media outlets, as both influence news coverage. 1594 S. HARLOW ET AL. RQ2a-d: How do language and location of media outlet predict use of the a) riot, b) confronta- tion, c) spectacle, and d) debate frames in protest-related articles shared on social media? Share-Ability and Online News Social media like Facebook and Twitter have amplified people’s ability to spread, shape, and comment on online news, and an increasing number of people retrieve news from social media (Matsa & Shearer, 2018). This shift demands an understanding of the charac- teristics of news coverage that is shared on social media. Social media users’ capability to overcome mainstream media’s gatekeepers through their recommendations and personal networks is one of the main disruptions to the media ecosystem (Tenenboim and Cohen 2015). Social recommendations raise questions about what elements make news spread- able in digital spaces. While some studies focus on the psychological processes (e.g., Ho and Dempsey 2010) or socio-technical affordances (e.g., Carlson 2015) behind online engagement, a growing line of studies has focused on how content features trigger share-ability (García-Perdomo et al. 2018; Trilling, Tolochko, and Burscher 2017); this is the perspective taken in the present research inquiry. News values, tone, and valence can influence audience reactions on social media (e.g., Bright 2016; García-Perdomo et al. 2018; Brown et al. 2018). Trilling, Tolochko, and Burscher (2017) found that conflict and human interest increased interactions. Surprise, controversy, and relevance likewise can predict news shared online (Rudat et al., 2014). This present study broadens the scope of previous research by exploring audience engagement with protester coverage on social media, posing the following research questions: RQ3: What media outlet characteristics, protest characteristics, and protest-paradigm cover- age components predict the number of Facebook interactions? RQ4: What media outlet characteristics, protest characteristics, and protest-paradigm cover- age components predict the number of Twitter interactions? Methods To better understand the journalistic coverage of protests by news media outlets globally, a content analysis of news stories shared on social media was conducted. Articles were collected using Newswhip’s Insights platform, which includes records from the application program interfaces (APIs) of over 50,000 news organizations worldwide, and archives all unique universal record locators (URL) for articles shared on social media. Each URL is then tracked within the APIs of social media networks to provide the number of times the URL (news article link) was engaged with on Facebook and Twitter. A search for all articles with the words “protest” and “protester” and their Spanish equivalents, “protesta,” “protestar,” and “protestante,” was conducted from Jan. 1, 2014 to Dec. 31, 2014. Due to limited download capacity, the top 24,000 most-shared articles from Facebook in English (n = 12,000) and Spanish (12,000) were collected. Though these selection criteria were not without limitation, Facebook shares were prioritized because its network was significantly larger and thus more representative from a global perspective. Facebook data available included the total number of likes, shares, and com- ments that appeared cumulatively on all public links shared. Twitter data available  JOURNALISM STUDIES 1595 included the total number of times a link appeared as an original tweet, retweet, or any tweets from link-shortening sites. Favorites on Twitter were not available. From the total 24,000 posts, a random sample of 750 English articles and 750 Spanish articles was ana- lyzed. A total of 62 articles were not relevant, resulting in a final sample of 692 English articles and 746 Spanish articles. Dead hyperlinks were replaced. All four authors of this study served as coders. Intercoder reliability was calculated on 10% of the sample, resulting in Krippendorff’s alphas that ranged from.71 to 1.0. Variables’ individual alpha levels and operationalizations are presented below. News Organization Variables Coders identified the type of news organization as online versions of traditional/mainstream news outlets (e.g., nytimes.com, CBS, CNN, FOX, BBC or any local newspaper/television station); alternative media outlet/activist/partisan website (sites that self-identify as such, i.e., Democracy Now, Alternet, Anonymous, Latino Rebels, Radio Free Europe); and online-native media outlets (online-native news sites, social news aggregators, video sharing sites, portal news sites, online media outlets with no explicit alternative desig- nation in their “about” sections). Coders were instructed to visit each website’s “about” section to identify the outlet’s type (α =.79). Dummy variables were created for media type, with traditional/mainstream as the reference. Coders identified the geographic region of the news outlet from seven regions: Central and South America & Caribbean; Europe; the United States and Canada; Africa; Asia; Middle East; and the Pacific Islands of Australia or New Zealand (α =.77). Due to low fre- quencies, Pacific was collapsed with Asia. Dummy variables were created for media region, with U.S./Canada as the reference. Protest Identification Variables Coders identified the geographic region of the protest: Central and South America & Carib- bean; Europe; United States and Canada; Africa; Asia; Middle East; Pacific (Australia or New Zealand); more than one region/worldwide; or not applicable (α =.85). Pacific again was collapsed into Asia. Dummy variables were created for protest region with the U.S./ Canada as the reference. Additionally, using the geographic location for the media outlet and the protest, variables were re-coded into whether the media outlet was located in the same region as the location of the protest. The type of protest was open coded. Coders wrote descriptions including what the protest was about, the aim of the protest, whether any names of people were mentioned as symbolizing the protest, and whether specific hashtags were mentioned. The authors convened several times to discuss the categories and thus increase interpretive validity (Lindlof and Taylor 2002). Researchers together settled on 33 specific types of protests (i.e., Black Lives Matter, pro-freedom of expression, indigenous rights) that ultimately were collapsed into six broad categories: (1) anti-government/corruption, (2) human rights/justice/peace, (3) socio-economic (i.e., labor issues, net neutrality, education), (4) environment/animals, (5) religious (i.e., pro-Islam, pro-Palestine), and (6) conservative/ revivalist (i.e., anti-abortion, anti-LGBTQ+ rights, anti-immigration, pro-gun rights). The first five categories represent progressive protests or those that challenge the status 1596 S. HARLOW ET AL. quo, while the final category includes those protests that seek to uphold the status quo or maintain hegemonic power structures. Dummy variables were created for the protest type, with anti-government/corruption as the reference. Protest Frames Coders identified use of the protest paradigm’s riot, confrontation, spectacle, or debate frame. Frames were not mutually exclusive. The riot frame included a focus on the violence of protestors through rioting, looting, or causing damage to public property or society (α =.83). Confrontation included a focus on clashes between protesters and police or auth- orities, or the arrests of protesters (α =.81). The spectacle frame included articles that focused on the emotions, drama, or unusualness of protests (α =.80). The debate frame focused on the social critique of the movement, characterized by the presence of protes- ters’ viewpoints and demands. It also included attention to the background and history of a movement (α =.80). Devices Coders identified if the article attributed violence or violent acts to protesters (α =.73). Mentions of peaceful protests also were coded (α =.71). Sourcing Coders counted the number of named individuals directly quoted in each story. Consider- ing all the sources used in the story, coders identified whether protesters or officials (gov- ernment officials, official representatives of the organization being protested, or police) were cited more often, equally, or not at all (α =.85). Social Media Interactions Variables Two social media interaction variables were used in this study. For Facebook, the numbers of likes, shares, and comments were combined into one composite interaction variable. On Twitter, only cumulative retweet counts were available via Newswhip, so interactions were limited to the number of retweets on Twitter. These outcome variables were normalized using a log-10 transformation so that outcomes would provide more appropriate results for use in linear regressions. Results Data Overview Most online articles analyzed were published on the websites of mainstream news outlets (43.3%), followed by alternative outlets (32.1%) and online-native sites (24.6%). The most common regions where outlets were located were U.S./Canada (40.7%), Latin America/Car- ibbean (33.1%), Europe (19%), Asia/Pacific (4%), Middle East (1.9%), and Africa (1.5%). Results showed that most articles were about protests related to human rights/justice/ peace (43.8%), followed by anti-government/corruption (35.5%), socio-economic issues  JOURNALISM STUDIES 1597 (8.2%), conservative/revivalist (6.5%), environment/animals (4.3%), and religious (1.7%). Most stories were about protests located in the U.S./Canada (36.6%), followed by Latin America/Caribbean (29.3%), Europe (16.6%), Asia/Pacific (6.6%), Middle East (4.4%), mul- tiple regions (3.7%), and Africa (2.7%). Nearly two-thirds of stories (64.5%) were published in media outlets in the same regions where the protests occurred, whereas 34.6% were about protests taking place in regions other than where the media outlet was located. For frames, 56.4% of stories contained a debate frame, 45.6% a confrontation frame, 34.1% a spectacle frame, and 21.2% a riot frame. Stories that quoted protesters the most totaled 34.6%, while 22.3% quoted officials the most, 9.4% quoted protesters and officials equally, and 33.7% quoted neither protesters nor officials. More articles men- tioned violence (23.1%) than peacefulness (14.1%). Protest Paradigm Predictors RQ1a-d and RQ2 a-d considered 1) how protest type and location and 2) media outlet language and location might predict adherence to the framing components of the protest paradigm. Four binary logistical regressions were run with the a) riot, b) confrontation, c) spectacle, and d) debate frames as the dependent variables (Tables 1–4). Results for RQ1a showed that articles about conservative protests had significantly decreased odds of having the riot frame [exp(B) =.143], and articles about socio-economic protests had significantly increased odds [exp(B) = 2.194] of having a riot frame. For RQ1b, stories about socio-economic protests [exp(B) =.227] and human rights/ justice/peace protests [exp(B) =.645], as well as those about protests in Europe [exp(B) =.521], had decreased odds of employing the confrontation frame. Table 1. Binary logistical regression predicting use of the riot frame Β S.E. eB Confrontation frame*** 1.095.217 2.991 Spectacle frame −.352.233.703 Debate frame*** −1.652.222.192 Media region: Lat Am/Car. −.160.376.852 Media region: Europe.075.367 1.078 Media region: Africa −.599 1.177.549 Media region: Asia/Pacific 1.243.646 3.465 Media region: Middle East.853.929 2.347 Media type: Online.129.253 1.137 Media type: alternative.172.245 1.187 Media language: English −.168.345.846 Protest type: Socio-economic*.786.381 2.194 Protest type: Human rights.338.270 1.402 Protest type: Envir/animals −.078.555.925 Protest type: Religious.411.743 1.508 Protest type: Conservative* −1.948.780.143 Protest region: Lat Am/Car..664.351 1.943 Protest region: Europe.390.422 1.477 Protest region: Africa 1.140.809 3.128 Protest region: Asia/Pacific −.863.606.422 Protest region: Middle East.338.655 1.402 Media region different protest region −.235.260.229 Constant** −1.476.502.229 *p