Characteristics of Protest Events in Television News Coverage (PDF)

FROM THE STREET TO THE SCREEN: CHARACTERISTICS OF PROTEST EVENTS AS DETERMINANTS OF TELEVISION NEWS COVERAGE* Ruud Wouters† Media attention is a crucial resource for demonstrators seeking to influence policy. This article assesses the determinants of television coverage for pro...

FROM THE STREET TO THE SCREEN: CHARACTERISTICS OF PROTEST EVENTS AS DETERMINANTS OF TELEVISION NEWS COVERAGE* Ruud Wouters† Media attention is a crucial resource for demonstrators seeking to influence policy. This article assesses the determinants of television coverage for protest events. Police archive data for the city of Brussels is compared to newscast data of the biggest public and commercial station in Belgium (2003-2010). Results show that few demonstrations pass the television gates (11%). Above all, protest size accounts for newsworthiness. Disruptive and symbolic actions also attract the media spotlight, confirming drama and visuals as critical television news values. Whereas symbolism matters for media presence, it does not hold for headline (prominence) or length of coverage (volume). New social movements are especially likely to stage symbolic actions. As a consequence, their presence on the screen is less a function of their numeric weight in the street. Distinctive selection mechanisms are at work on the commercial and public broadcaster, sug- gesting that media ownership matters for news selection. Protest actions are communicative acts staged to signal situations of injustice. Not the act itself, but the reaction of others to the act is what brings about political leverage (Lipsky 1968). The power of protest therefore strongly depends on diffusion and amplification; protestors’ ability to gain access to means of mass communication is key. Media attention might raise the public saliency of an issue, might activate otherwise silent bystander publics and might put pressure on political elites to pay attention (Gamson 2004). Koopmans (2004) asserts that politicians only react to social movements if and as they are depicted in the media arena. Gaining media attention is easier said than done, however. News is to a far greater extent the result of a process of exclusion than inclusion. Numerous events happen every day and only few end up in pixel or print. In this article I scrutinize the selection of protest events in television news. Two straightforward questions take center stage: To what extent do protest actions become news? And, which characteristics of protest events make them particularly newsworthy? Previous studies tackling media selection of protest have focused primarily on newspaper cover- age. This focus on the press primarily stemmed from methodological concerns and resulted in a distinct selection-bias literature that probed the validity and reliability of newspaper data (Rucht, Koopmans, and Neidhardt 1998). However, regarding protest coverage, television news is at least as relevant a medium to study as newspapers. It is still the most important source of information about political affairs and society for the citizen public (Curran, Iyengar, Lund, and Salovaara-Moring 2009). The fact that television adds pictures to words makes it extremely well-suited to communicate the vibrant atmosphere of protest as well (Small 1994; Etzioni 1969). Not only do the visual stimuli of televised protests increase audience recall, the subsequent emotional arousal might be more effective than newspaper articles in pressuring viewers to choose sides (Graber 1990, 2001). Moreover, because of these visual conditions, other selection mechanisms may be at work in television news when compared to newspapers. The chance to secure interesting video footage, for instance, might be a more important news _______________________________ * The author would like to thank Stefaan Walgrave, Shanto Iyengar, the M²P research group and two anonymous reviewers for their helpful feedback and suggestions on earlier drafts of this article. † Ruud Wouters is a PhD. student in the department of Political Science at the University of Antwerp, and a member of the Media, Movements and Politics Research group (www.m2p.be). Please direct correspondence to Ruud Wouters, University of Antwerp, Sint-Jacobstraat 2, 2000 Antwerp, Belgium. E-mail: [email protected]. © 2013 Mobilization: An International Journal 18(1): 83-105 84 Mobilization value for television (Harcup and O'Neill 2001). In sum, television news has a far greater reach and a far greater potential influence compared to newspapers. If one is less interested in the methodological issue of data-source bias and more interested in the potential impact of protest, one needs to focus on the selection mechanisms of television news coverage (Walgrave and Vliegenthart 2012 ; Smidt 2012). Besides the focus on televised protest, this article’s second contribution relates to the operationalization of media attention. Most previous studies define media selection simply as a binary condition of whether an event succeeded in attracting coverage or not. Media coverage of demonstrations varies significantly, however. Whereas some events are treated in depth, other protests are given limited airtime in so-called “news carrousels.”1 In the same way, some protests may be featured more prominently than others. When journalists assign headline status to an event, they clearly acknowledge its importance, a judgment that is not without its public consequence. Behr and Iyengar (1985) found that lead news stories are especially likely to set the public agenda. So, besides the presence of protest in a particular newscast, I also take the volume and prominence of protest coverage into account. In order to tackle the central questions of this study, I utilize two databases. Police record data of all protest events organized in Brussels (2003-2010) are considered as the universe of events. Such extra-media data allow for the most robust test of media selection (Hocke 1998; Oliver and Myers 1999). Content analysis data of the flagship newscasts of Belgium’s biggest commercial station and the Belgian public broadcasting service represent the media filter.2 The article’s third contribution lies in the assessment of selection mechanisms across two different stations characterized by different ownership structures. Results show television coverage to be a numbers game. Mass mobilization not only results in higher chances of media coverage; large protest events also more easily make it into the headlines and are given more airtime. The same rationale holds for disruptive actions. Whereas symbolic actions and actions organized by more professionalized organizations more easily become news, they are not more likely to be prominent or lengthy news items. Protest events staged by migrant organizations, on the other hand, received significantly less airtime. Interestingly, new social movements are especially likely to stage symbolic actions. It is these colorful and camera-ready tactics, more than the size of the demonstrations, that lead to greater presence on the screen. Media ownership, finally, matters to a great extent for the selection of protest events: the news desk of the public broad- casting service (VRT) covers many more protest events, whereas the only commercial station with a newscast (VTM) is more confined to events that score high on disruption and drama. MASS MEDIA, SOCIAL MOVEMENTS, AND PROTEST EVENTS Mass media attention is an important resource for movements seeking to influence the polity. Media attention validates a movement and is crucial to the mobilization processes (Walgrave and Manssens 2005) and membership recruitment (Vliegenthart, Oegema, and Klandermans 2005). Media coverage is an asset that helps to enlarge the scope of a given conflict (Gamson and Wolfsfeld 1993). Although social movements need media attention more than institutional political actors, the literature suggests that movements have to fight an uphill battle to gain media attention (Molotch 1979; Rohlinger 2006; Wolfsfeld 1997; Ryan 1991). Two broad strands of literature account for access to the mass media arena: one stream focuses on the production logic of journalism (news routines) and another stream focuses on event characteristics (news values). In both perspectives, journalists act as gatekeepers, deciding what becomes news and what does not. Research on news routines looks at the day-to-day practicalities of the newsroom to explain media selection. Besides media ownership and the importance of profit making as overarching influences on newsroom dynamics, one of the most important routines in journalism is the assignment of reporters to “beats” (Shoemaker and Reese 1996). A beat is both a specific topic (crime, politics) and a specific location (court house, From the Street to the Screen 85 parliament). Beats assure a steady stream of coverage. Reflecting the mass media’s role of fourth estate, the most important news beat is the political beat. As a consequence, the news- gathering apparatus’ day-to-day attention is far more focused on institutional sources (Bennett 1990; Tresh 2009). Whereas political elites are caught in the newsgathering net whether they want it or not, social movements must more actively attract the attention of the media spotlight. One way to obtain coverage is to anticipate these routines and facilitate the newsgathering process. The fact that journalists work under clear budgetary constraints and with strict dead- lines suggests news production needs to be efficient (Tuchman 1973, 1978). The work of journalists is made easier when groups present information in timely press releases and put forward leaders and spokespersons who speak in short and quotable sound bites. Social movements willing to enter the mass media arena must adapt to such news routines (Schudson 1996). Andrews and Caren (2010) confirm that environmental groups that send more press releases, maintain contact with reporters, and call newspapers about their activities, are more likely to make it into the news. Sobieraj (2010), however, questions the effectiveness of adaptation to news routines. Opposite norms apply for political insiders and outsiders. Activists who try to be professional, quotable and credible do not fit the schema journalists have of political-outsider news sources and therefore will be ignored. According to Sobieraj, good activist sources, first of all, need to be authentic: journalists prefer them to possess an amateur status and moral authority, as well as having personal proximity to the issue at hand. Profession- alization in this respect does exactly the opposite: it lowers the attractiveness and news- worthiness of protestors. News value theory, on the other hand, considers newsworthiness as a product of event char- acteristics (Harcup and O'Neill 2001; Galtung and Ruge 1965). Events that score high on any number of news factors are more likely to elicit media attention. Such event characteristics are, for instance, unambiguousness, unexpectedness, possibilities of personalization and drama- tization, novelty, the involvement of elites, negativity, and conflict (Tresh 2009). These news values are not fixed in the sense that certain combinations guarantee media resonance. They vary between broadcasting systems and media outlets (Cushion 2010). In general, for resource- poor social movements, courting the media spotlight is less about news routines and more about emphasizing event characteristics known to have news value. Challengers succeed in entering the media arena only by carrying out exceptional acts (Wolfsfeld 1997). Danielian and Page (1994), for instance, hold that protest actions are the most surefire way for citizen groups to make it into the news. Andrews and Caren (2010) similarly conclude that using the outsider tactic of protest enhances the likelihood of media attention. In sum, protest tactics clearly matter for social movements that want to make their claims public. Yet not every protest action is equally likely to become news. Specific characteristics of protest events will increase or decrease the likeliness of coverage. MEDIA SELECTION OF PROTEST EVENTS AND ITS DETERMINANTS Social movement scholars have developed a long-standing interest in the selection of protest events by mass media. Interest in the validity and reliability of newspaper data sparked a distinct selection bias literature (Earl, Martin, McCarthy, and Soule 2004; Danzger 1975; Snyder and Kelly 1977; Mueller 1997; Oritz, Myers, Walls, and Diaz 2005; Myers and Caniglia 2004). First testing for intermedia selection, an important improvement was made by relying on media- independent data, found in police archives (Barranco and Wisler 1999; Fillieule 1997, 1998; Hocke 1998; McCarthy, McPhail, and Smith 1996; McCarthy, Titarenko, McPhail, Rafail, and Augustyn 2008; Oliver and Maney 2000; Oliver and Myers 1999). Table 1 gives an overview of those studies that explicitly tackled the problem of protest-event selection by relying on police archive data. The overview shows that a great variety of cases have been looked into. Small local cities, as well as national capitals have been studied, in Europe as well as the United States. 86 Mobilization Table 1. Overview of Studies Assessing Media Selection Using Police Archive Data a Author City Period N Determinants Media Outlet Coverage McCarthy, McPhail, USA 1 local and and Smith (1996) -Washington Size LN: 6.9% 1 national DC 1982, 1991 3,065 Issue NN: 3.0% newspaper, Newshole TV: 2.1% Television news Fillieule (1997) France - Paris January- 2 national - Nantes 758 Size NN: 9.6% June 1989 newspapers - Marseilles Hocke (1998, 2000) Germany Size 1 local and 2 - Freiburg LN: 37.8% 1983-1989 196 Disruption national NN: 4.6% Sponsor newspapers Barranco and Wisler Switzerland Size (1999) - Bern 1965-1994 Disruption 3 local and 1 NN: 51.2% - Geneva 1965-1994 1,538 Sponsor national LN: 63% - Zurich 1965-1991 Newshole newspaper - Basel 1981, 1988-1994 Oliver and Maney USA Size (2000) b - Madison, Disruption 2 local WI 1993-1996 541 LN: 46% Sponsor newspapers Newshole McCarthy, Titarenko, Belarus Size McPhail, Rafail, - Minsk 1990-1995 817 Disruption 4 ideologically and Augustyn (2008) different AN: 30.6% Sponsor newspapers Issue a b Notes: LN = local newspaper; NN = national newspaper; AN = all newspapers, TV = television news. See also Oliver and Myers (1999) for an analysis of a single year of the same dataset. Both local and national newspapers have been chosen as media filters. Strikingly, however, only one study (McCarthy et al. 1996) assessed selection for television news. What information does table 1 provide about the degree of selectivity? Percentages of media coverage are shown to vary greatly across media outlets and are clearly classified by the scope of the media outlet, with local newspapers systematically covering more protest events than national ones, and television news—with a far more limited newshole—covering fewer protest events than newspapers.3 Variety within same scope of media outlets nevertheless remains high. Besides scope of the media outlet,4 the ideological stance of the media outlet (Mueller 1997; McCarthy et al. 2008) is also considered as a determinant of selectivity. In the current study, both television stations are based in Brussels, are national in scope and have a very similar newshole.5 Ideological stances in broadcasting are, contrary to the American context, atypical in Northern Europe (Hallin and Mancini 2004). Another potentially meaningful classification, however, is the distinction between commercial and public television stations. Commercial players were allowed to enter European television markets only since the late 1980s and early 1990s (Holtz-Bacha and Norris 2001). Despite the great similarities between both stations, I expect the ownership model of both stations to result in different patterns of media selection. Whereas public broadcasters focus more on “hard” political and institutional news, commercial stations tend to prefer “soft” news with a more sensationalist and entertainment- oriented focus in what they cover and how they cover it (Gamson, Croteau, Hoynes, and Sasson 1992; De Swert 2007; Reinemann, Stanyer, Scherr, and Legnante 2012). As protest events typically tackle policy-related issues, I expect the hard news focus of the public broadcaster to act as a window of opportunity for protest events. I therefore expect the public broadcaster to cover more protest events compared to the commercial channel. Although protest events From the Street to the Screen 87 typically tackle hard-news issues, the form of protest (colorful and vivid, with participants singing, dancing and yelling) clearly fits soft-news criteria. I expect the soft-news focus of the commercial station will result in a distinct selection pattern: the news values of drama, disruption, spectacle, and conflict—news values that tie in with sensationalism and entertain- ment—would in this regard be stronger determinants of media selection on the private than on the public station. As a consequence, I expect protest events covered by the commercial station to more frequently receive a headline status and to get more airtime compared to protest coverage on the public channel. If commercial stations cover fewer, and more spectacular, events, it is logical to expect that commercial television will treat protest more prominently and at greater length. In the opposite direction, I expect that protest events that succeed in eliciting attention from both stations, will be covered more prominently and at geater length by the public broadcaster. These expectations result in the following hypotheses: H1: The public broadcaster, with its focus on hard news, covers more protest events than the commercial broadcaster. H2: Protest selection by the commercial channel is characterized by a more sensationalistic selection mechanism than protest selection by the public broadcaster. H3a: Protest events are more likely to make headline news on the commercial channel compared to protest events on the public channel. H3b: Protest events covered by both stations are less likely to make headline news on the commercial channel. H4a: Protest events get more airtime on the commercial channel compared to protest event on the public channel. H4b: Protest events covered by both stations get less airtime on the commercial channel compared to the public channel. Besides the extent of media coverage for protest events, this study also sets out to tackle the determinants of media coverage. News selection of protest is by no means arbitrary, and a range of determinants influences the likelihood of coverage. Della Porta and Diani (1999) distinguish three different logics of protest: a logic of numbers, disruption, and bearing witness. These three logics of collective action clearly relate to criteria of newsworthiness: they hold characteristics explicitly correspond to the news values known from journalism studies. The logic of numbers contends that large crowds demonstrate the social power of a group (Tilly 2004). Mass mobilizations raise the chances that journalists will judge the event as relevant and legitimate. Demonstration size proves to be the strongest and most consistent factor predicting newspaper coverage of protest events (McCarthy et al. 1996; Oliver and Maney 2000; Oliver and Meyers 1999; Hocke 1998; Fillieule 1997). In the same vein, I expect television coverage of protest to be largely a function of demonstration size. H5a: Protest events that succeed in drawing large numbers of participants are more likely to become news. H5b: Protest events that succeed in drawing larger numbers of participants get more airtime and will more frequently be mentioned as headlines. Besides the logic of numbers, della Porta and Diani (1999) also discern a logic of material damage. Protest organizers may aim at large numbers, but many potential protestors may drop out (Klandermans 1997; Van Laer 2011) and demonstration turnout is unsure. De Nardo (1985) suggests that groups that cannot lean on huge constituencies have to resort to more extreme and 88 Mobilization confrontational tactics. Disruption and violence clearly match with the media’s need for drama and conflict. From a news-value perspective such events should be more likely to result in news coverage. With the normalization and routinization of protest, the presence of explicitly violent protest has declined, however (Van Aelst and Walgrave 2001). Koopmans (1993) therefore speaks of confrontational protest, in the sense of acts of “civil disobedience.” Not material damage, but getting arrested—with resulting captivating protagonist-antagonist images— becomes the activists’ explicit goal. Here, we test for such a logic of disruption. Barranco and Wisler (1999) and McCarthy et al. (2008) found that events with arrests were more likely to make it into the papers, whereas Oliver and Myers (1999) and Oliver and Maney (2000) did not find such a disorder dynamic to be at work. Myers and Caniglia (2004) found that within a group of civil disorders reported in local newspapers, the more intense disorders (deaths, arrests, arson, duration) more easily made it into the national press. Here I expect violence and disorder to be consistent and strong news values. Television is by definition a far more visual medium than the printed press, and as violence and disruption clearly fit the need of television for spectacular images and conflict, I believe disruption to be a strong predictor of newsworthiness. Moreover, when it comes to media prominence and volume, I expect disruption to have a strong and positive effect. Besides overt conflict, unexpected protest can also be judged as disruptive. For example, protest events that are carried out without a legal permit can catch authorities off-guard and are typically more disruptive of public life. Contrary to standard news-value expectations, these kinds of unexpected protest events are less likely to receive media coverage. Both illegal demonstrations in Switzerland (Barranco and Wisler 1999) and unpermitted protests in Madison (Oliver and Myers 1999) seem to catch both police and journalists by surprise. H6a: Disruptive protest events are more likely to become news. H6b: Disruptive protest events get more airtime and will more frequently be mentioned as headlines. H7: Protest events that are unexpected are less likely to become news. The logic of bearing witness, finally, fits groups that cannot or choose not to play on numerical strength or disruption. Here, persuasion is obtained via symbolically enacting one’s claim. Imagination, creativity, and sensitivity to values and culture are key (Rochon 1988). Whereas della Porta and Diani (1999) consider civil disobedience, which involves high personal commitment and risk of being arrested, to be a first aspect of the logic of bearing witness, my definition is more stringent and moves the symbolic dimension to center stage. The logic of bearing witness consists of actions that are theatrical happenings, carefully scripted and performed, in order to respond to mass media criteria of news judgment (Pichardo 1997; Tarrow 1994; Keck and Sikkink 1998). Examples of such symbolic actions include Amnesty Inter- national calling for the shutdown of Guantanamo by wearing orange overalls and imitating tor- ture activities, international aid organizations highlighting the plight of refugees by installing refugee tent camps in city centers, or farmers publicly dumping milk to protest low milk prices. These events are not events of mass mobilization. They can be disruptive, but do not have to be. Their greatest common denominator is that they are public performances in the form of pseudo– events. As such, they follow the rules of the media game and use key symbols to draw media attention. In their Visualizing Deviance, Ericson, Baranek and Chan (1987, 274-275) describe how a citizen group opposed to nuclear weapons set up a symbolic action: The source organization selected a key location (a plant involved in weapons manufacture); a good day in summer when news was likely to be ‘slow’ because key source bureaucracies were in recess; and, sent a release, along with promises via telephone calls, that there would be ‘violence’ of some sort. One television station had three crews on location. The source organi- From the Street to the Screen 89 zation arranged for five people to jump over a fence to throw red paint, signifying blood, on a wall. They were stopped and arrested for trespass or public mischief. Meanwhile two women were able to paint with stencils six green doves on a wall before they were arrested.... The source organization had also come equipped with a variety of props—including banners (e.g., ‘Ban the Bomb’) and a coffin—all of which were focused on by cameramen, along with shots of the paint on the wall and the persons responsible being arrested. The reporters recognized they were simply following the script of their source, and one remarked while editing that this item was a prime example of creating the news. It is the scripted nature, the staged performance, the reliance on news-routine knowledge, and the visually attractive aspect that make these symbolic actions particularly suitable for television coverage. Symbolic actions carry this media logic (Altheide and Snow 1979) to its extremes. This is why I expect symbolic actions are more likely to resonate with television journalists’ conceptions of what constitutes news.6 There is a price, however, that comes with the fact that symbolic actions tend to so closely follow this media logic. Scholars of political communication often describe the relationship between journalists and their sources as a dance, with both partners needing each other (publicity vs. information), and with a constant struggle over power (who leads the dance?). In the words of Strömbäck and Nord (2006, 161), “On the dance floor, political actors are doing what they can to invite the journalists to dance, but ultimately, it are [sic] the journalists who choose who they are going to dance with.” I expect that activists staging events that follow the logic of bearing witness succeed in convincing journalists to cover events that otherwise would not have been judged newsworthy. However, as a consequence, the chance that these events will be featured prominently and at length will be lower. This leads to the following hypotheses: H8: Protest events that are symbolic are more likely to become news than are events that do not display such elements. H9: Protest events that are dramatically symbolic get less airtime and will be featured less prominently than events that do not display such elements. Besides these logics of protest, I believe protest form matters as well. I distinguish between static and dynamic protest, and expect static protest to be less newsworthy. There are several reasons why a protest that moves from point A to point B (i.e., marches) could be more news- worthy. For one, marches are more likely to draw big numbers, and also are more disruptive because of the traffic jams they cause. McCarthy et al. (1996) believe dynamic demonstrations are also more visible and audible in public life as they more explicitly and overtly show off by parading in the streets. Finally, the fact that marches are organized far more infrequently makes them extraordinary and thus newsworthy. On the same note, one could also expect static protest to be more immediately focused on appealing to bystanders at the event, and less focused on appealing to a mass audience. Clearly the static form of protest—especially activities focused on distributing literature—is less about making headlines and more about convincing passing pedestrians. The function of static protest is more often to inform than to persuade.7 For these reasons, I expect static protest events to pass through the media gates less easily. H10: Static protest events are less likely to become news than dynamic protest events. The logics and form of protest are not the only factors that matter for newsworthiness. News routines also matter. Events staged by more established and professionalized sponsoring organizations more easily pass the media gates (Oliver and Myers 1999; Hocke 1998; McCarthy et al. 2008). These organizations are more resourceful, can maintain relationships with jour- nalists, and can feed detailed information to the press (Andrews and Caren 2010). In sum, these groups can adapt to news routines, thereby facilitating the newsgathering process. These groups are also more likely to have media standing. That is to say, they have previously proven to be a 90 Mobilization legitimate (and perhaps trustworthy) news source, and this enhances their chances of making it into the news once again (Ferree, Gamson, Gerhards, and Rucht 2002). Although the distinction is not without its critics, a well-known and well-suited typology in this respect distinguishes between old and new social movements (Mellucci 1996; Williams 2004). The typology has several dimensions, including one measuring institutionalization and professionalization, with new social movements (NSM) having weaker organizational structures and being less institutionally embedded. The distinction is also issue related, however. Old social movements (OSM) care about redistributive, materialist bread-and-butter issues. New social movements—like the peace, environmental, and north-south movement—advocate post- materialist concerns (Buechler 2000; Verhulst 2011). With the increased internationalization of protest (della Porta and Tarrow 2005), I also add migrant and asylum seeker organizations as a distinct type of social movements, as these are generally less organized and have fewer resources than their domestic counterparts (Chabanet and Giugni 2008). Since OSMs tend to be more professionalized and institutionalized, I expect them to be better equipped to court the media. Especially in neocorporatist Belgium, typical old social movements like unions and professional organizations take strong insider positions in the policy making process (Martens, Van Gyes, and Van der Hallen 2002). With the media following the trail of political power, OSM protest hence can be expected to generate media attention more easily. Another dynamic might also be at work: OSM issues (labor market, economy, agriculture, social affairs) might be featured more frequently and prominently in the news, resulting in a favorable media climate for demonstrations tackling “old” issues. Not the agency or profession- alization of the organization, but rather the “discursive opportunity structure” of the media arena then accounts for differences in media presence. Regardless of which mechanism might be at work, this leads to the following hypotheses: H11a: Protest events organized by old social movements are more likely to become news than are events organized by new social movements. H11b: Protest events by new social movements are more likely to become news than are events organized by migrant and asylum seeker organization. Finally, this study takes into account one more typical newshole characteristic. Demonstra- tions compete with other events for media attention. On days when news is slow, chances for any event to get media access should increase. One could expect demonstrations to take advan- tage of the news vacuum that results from the lack of economic or political news over week- ends. This expectation is corroborated by Barranco and Wisler (1999), yet rejected by McCarthy et al. (1996) and Oliver and collaborators (1999, 2000). Less-populated news desks on week- ends (McCarthy et al. 1996) and, for Wisconsin, the lack of legislative activity on Mondays (Oliver et al. 1999, 2000), account for these findings. Whereas newspapers cover, at best, yesterday’s events, television news is a daily occurrence, broadcasting both on Saturday and Sunday. I therefore expect weekend events are more likely to become news than events staged on weekdays. H12: Protest events staged during weekends are more likely to become news than are protest events staged on weekdays RESEARCH DESIGN In order to assess the filtering mechanisms of mass media I compare data from police archives with media data (Rosengren 1974). The research period runs from January 2003 to December 2010. The setting of the study is Brussels or, more precisely, the police jurisdiction Brussel Hoofdstad-Elsene. Brussels is the Belgian capital as well as the political heart of both Belgium From the Street to the Screen 91 and Europe. The main political institutions of the different Belgian regions are located in this jurisdiction, as are the headquarters for many European agencies and foreign embassies. This specific context makes Brussels a particularly attractive location for protestors interested in showing their discontent. Police record data. Access was granted to the archives of the Brussels Capital Police District.8 In Brussels, every protest action needs permission. Copies of requests and permits were coded on-site by a group of MA students under supervision of the author. In an ideal situation, one report would give a start-to-finish perspective of a particular demonstration. The document would open with a letter of request to the mayor, would contain a risk analysis and police deployment plan, a short description of the protest issue, and an estimation of the number of participants, and the organizations involved. The document would end with a short follow-up report after the demonstration, describing the flow of events, the actual number of participants, and any irregularities. Such after-demonstration documents were only included for 62% of the protest events. Of those protest events that contained an after-demonstration report, 98% effec- tively took place.9 Since I include all demonstrations in the analysis, the expected number of participants is used as a measure of demonstration size, rather than the actual number of participants.10 If a demonstration was not requested, but discovered by patrolling police officers, a short after-demonstration report was written by the patrolling officers and added to the archive. These unpermitted demonstrations comprise 5.5% of the police record database and are also included in the analysis. Of course, police officers are not social scientists, and many prefer action on the street to administrative deskwork. Many demonstrations found in the media database were not present in the police archive. I am nevertheless convinced that the database is rather complete.11 The police- record database covers many more protest events than the media database, and, as the goal of this article is not to draw a perfect picture of the real protest world but to investigate the media filtering mechanisms, the data make it possible to examine the gap between police records and media coverage. Besides protest events, the archive contains other events related to public police inter- ventions (police disposition in relation to football games, foreign head-of-state visits, royal family member actes de présence).12 The unit of analysis in this study is the protest event, and partly follows Fillieule’s criteria for selecting protest events. Fillieule (1998) uses the number of participants, a publicly expressive dimension, a political or social nature, the nature of the organizers, and the form of the event as dimensions for the selection of events as protest events. As social movement organizations do not have a monopoly on protest strategies, no organi- zational criteria were imposed. Mere fundraising actions of social movements organizations were excluded. Media data. The University of Antwerp and Leuven host the Electronic News Archive (www.ena.be). The ENA is a census dataset that daily records and codes the 19 o’clock flagship newscasts of both the biggest commercial (VTM) and public broadcaster (VRT). The database runs from 2003 to the present. The coding scheme is elaborated at both the item and actor level. The Belgian mass media market is a typical example of a duopoly. Contrary to trends in other countries, the audience for news is high and growing, with both 19 o’clock newscasts reaching on average 75% of the market during the period of interest.13 The ENA has no standard protest variable in its coding scheme. In order to trace news items about Brussels’ protest, a “needle in the haystack” word-search was performed for the words “protest,” its synonyms and its different notations.14 The search resulted in 12,411 protest news items (11.8% of all news items during the 2003-2010 period mentioned protest, of which Brussels protest is a subsample). The news items were manually linked to the police record database. A particular media report was linked to a particular demonstration drawn from the police archive only if date, topic, and protest location (Brussels) fully matched.15 Like previous research, the first dependent variable is a simple dummy variable reflecting whether a protest received media exposure or not. The study also takes into account the 92 Mobilization prominence of the protest coverage (whether the protest made it into the headlines of the newscast) and the volume of coverage (the duration of the total news item(s) about the protest event). Headline coverage is operationalized in the ENA archive as all items that get a short preview at the beginning of the newscast. Specific operational definitions of variables can be found in the Appendix. In order to test the impact of the different input variables, logistic regression is used when analyzing the dichotomous selection and prominence measures. As the volume of protest coverage is a count variable with the variance exceeding the mean, negative binomial regression is used. Given the discussion about stability of media selection (Ortiz et al. 2005), demonstration year is added as a control variable. RESULTS: PASSING THE MEDIA BARRICADES We coded protest events occurring between January 2003 and December 2010, 4,582 from police archive data. Brussels is flooded by protest events: every three out of four days at least one protest event takes place. On these demonstration days, on average two demonstrations are held; on 60 percent of the days with a demonstration, two or more actions occur. Table 2 sheds light on the demonstration context of Brussels. Most protests in Brussels are small and static. Six in ten protest events draw one hundred or fewer people to the streets and 85 percent of the protest events are static demonstrations. The distribution of demonstrators across events is extremely skewed: a small percentage of the protests (five percent) accounted for seventy-five percent of the demonstrators in the data. Migrant organizations and organizations of asylum seekers demonstrate most frequently. This confirms Table 2. Distribution of independent variables for protest events in Brussels, 2003-2010 % of % of N events demonstrators Logic of Numbers Less than 26 1,092 23.8 0.8 26 - 100 1,810 39.5 7.0 101 - 500 835 18.2 12.3 501 -1000 149 3.3 7.2 1001 - 5000 137 3.0 20 5000 - x 57 1.2 52.6 Logic of Disruption Disruption 202 4.4 17.7 Anticipated Disruption 804 17.5 57.7 Unexpected 251 5.5 0.9 Logic of Bearing Witness Symbolic Actions 410 8.9 7.4 Sponsoring Organization Old Social Movement 1,049 22.9 49.8 New Social Movement 811 17.7 12.9 Migrant and Asylum Seekers 2,058 44.9 27.0 Other 664 14.5 10.3 Demonstration Timing Weekend 962 21.0 41.9 Demonstration Form Static Actions 3,849 84.7 28.4 Demonstration Year 2003 372 8.1 15.7 2004 433 9.5 13.5 2005 564 12.3 16.9 2006 655 14.3 10.3 2007 585 12.8 11.2 2008 672 14.7 6.2 2009 679 14.8 11.3 2010 622 13.6 14.9 Notes: N= 4,582; In 89% of all demonstrations, an indication of demonstration size was incorporated in the police document. From the Street to the Screen 93 the image of protest as a weapon of the weak. Old social movements, like unions, welfare organ- izations, and professional organizations stage half as many demonstrations yet draw twice as many participants to the streets. New social movements organize the smallest actions, and have a distinct action repertoire: 50 percent of all symbolic actions are organized by new social movements. The “other” organization category, finally, includes demonstrations without a formal organizer, events staged by local organizations, or events staged by anarchists. The demonstration context of Brussels is largely peaceful. About five percent of the protest events are disruptive in nature. Disruption was operationalized as whether there were any arrests or any use of violence by the demonstrators (material damage, throwing eggs, setting garbage cans or cars on fire, blocking traffic, and wounded demonstrators, police officers, or bystanders). I also introduced a measure of anticipated disruption. In the run-up to an event, police forces perform a risk analysis and decide whether the deployment of a water cannon is necessary. If the demonstration atmosphere is expected to be especially contentious, a water cannon will be stationed to disrupt and disband demonstrators. The anticipated disruption variable also includes a measure of whether a counterdemonstration is expected. Water cannons were reserved in one out of every six demonstrations. One action in ten was considered to be a symbolic event. Unexpected events are those events where patrolling police officers caught the activists in the act, without a demonstration permit. Television coverage of protest How many protest events succeeded in making it from the street to the screen? How many make the headlines? What is the average duration of a protest item? As expected, many protests are organized, but few end up on the screen: only 11 percent (497 protest events) of all organ- ized protests ends up in the evening newscast on at least one of both television stations. Journalists show only the tip of the iceberg (Hocke 1998) as the vast majority of protest events were ignored. Table 3 shows the details. In line with expectations, the public broadcaster (VRT) is more sensitive to protest. It covers many more protest events. About half of the demonstrations covered by the public broad- caster do not make it onto the commercial channel. Both media outlets clearly differ in their propensity to cover protest, with the selection threshold for protest on the public channel being considerably lower. Hypothesis 1 is confirmed. Not only does the public broadcaster cover more protest events, actions that succeed in entering public TV are also featured more prominently.16 When only those demonstrations covered by both stations are taken into account, the result is even stronger: the public channel gives headline status almost twice as much compared to the private channel. Hypothesis 3a is therefore rejected and H3b is confirmed. Exactly how much exposure do protest actions generate? The average attention across all demonstrations on both stations is 150 seconds; 107 seconds if the amount of attention on both stations to the same particular event is not added up. The public and commercial stations do not Table 3. Presence, Prominence, and Volume of Media Coverage of Protest Events in Brussels Any Medium VRT Coverage (Public) VTM Coverage (Commercial) All Unique Both All Unique Both N of protest events 497 414 218 196 279 83 196 % of protest events 10.8 9.0 52.7 47.3 6.1 29.7 70.3 Total N 4,582 4,582 414 4,582 279 Prominence (%) 32.6 34.3 20.6 49.5 21.9 8.2 27.8 Volume (sec.) 150 107 76 142 108 77 122 Notes: N= 4,582; Unique coverage points towards the amount of demonstrations covered by a particular station that did not appear on the other station, both refers to demonstrations that made it in both newscasts. 94 Mobilization differ significantly in the amount of exposure they provide when all demonstrations are taken into account. Yet in those cases where both stations cover the same demonstration, it is the public station that covers protest more extensively.17 This is remarkable, as the average news item on the public station (97 sec.) is somewhat shorter compared to the average commercial news item (101 sec.). Hypothesis 4a is rejected while hypothesis 4b is confirmed. The bottom line is clear: not only does the public station cover more protest events, it features protest actions more prominently and more extensively. Determinants of Coverage Why are some events covered and others not? Or, formulated differently: which event characteristics succeed in securing the media spotlight? Figure 1 gives a first clue. The ability of protest to break through the media gates varies strongly across type of organization, and is clearly a function of demonstration size, but not protest frequency. For every type of organi- zation, event size closely mirrors media presence. Migrant organizations stage many small demonstrations, yet few are deemed newsworthy. Old social movements on the other hand, protest far less frequently but in far greater numbers in attendance. Whereas migrant organi- zations far more frequently take to the street, old social movements are far more prominent on the screen. Journalists’ perceptions of newsworthiness seem particularly sensitive to large demonstrations. Interestingly, size seems to play a far less important role when it comes to media selection of new social movement events. As shown by the bars in Figure 1, there is a large discrepancy between the number of demonstrators at NSM protest and their representation in the mass media arena. Whereas for all organization types demonstration size overestimates media coverage, NSMs appear more frequently than the size of their actions would suggest. Current literature suggests that NSMs are especially eager to stage symbolic actions, a finding confirmed by data on protest events in Brussels. The descriptives also show that these are rather small demon- strations.18 In order to specifically tease out the determinants of media selection, table 4 shows the outcomes of several binary logistic regressions and gauges net effects of the different independent variables. Figure 1. Distribution of Real World Protest Event Frequency Note: (N=4,582), mediated protest event frequency (N=497) and real world protest event frequency weighed by the number of expected participants, across four types of organizations. From the Street to the Screen 95 As indicated by figure 1, demonstration size is an important predictor of mass media cover- age, corroborating hypothesis 5a. Protestors who want to pass the media gates must aim for high numbers. If only demonstration size is taken into account, the pseudo R-square is.174. By in- cluding the categorical measure of demonstration size presented in the descriptives section, we can get a better grip on the effect of demonstration size. Both predicted probabilities and odds ratios constantly rise over the different categories of demonstration size. When all other vari- ables are kept at their means, demonstrations with less than 26 participants have a 4.3 percent chance of receiving media coverage, whereas demonstrations with more than 5,000 participants succeed in passing the media gates 51 percent of the time. Bivariate results obviously result in even stronger findings. 77% of the largest demonstrations (more than 5,000 participants) make it into the news. Media coverage is not only a numbers game, however. Adding other variables significantly improves the model. Besides size, the organizer or sponsor of the demonstration is also important in predicting media coverage. Who demonstrates clearly matters, with protests staged by old social move- ments more easily making it onto the screen than new social movements (H11a), and with NSM being more often singled out as newsworthy than migrant protests (H11b). Whereas evidence for the first is only marginally significant (p = 0.072), evidence for the latter is strongly significant. Two mechanisms can account for this finding. The first argument is one of profes- sionalization. Migrant organizations, asylum seekers, and the organizations grouped under the “other” umbrella (local organizations, demonstrations without formal organizer, anarchists) are definitely the least resourceful and least institutionalized organizations. They are in a far less Table 4. Coefficients of Logistic (Prominence) and Negative Binomial Regression (Volume) for Protest Events that Appeared in at Least One Newscast Prominence Volume B S.E. B S.E. Logic of Numbers Size 0.421*** 0.076 0.156*** 0.023 Logic of Disruption Disruption 0.646* 0.289 0.296** 0.101 Anticipated Disruption -0.082 0.255 0.065 0.088 Unexpected 0.914 0.470 0.246 0.165 Logic of Bearing Witness Symbolic Action -0.166 0.328 0.031 0.103 Sponsoring Org OSM 0.175 0.436 -0.056 0.143 (ref = other) NSM 0.366 0.465 -0.157 0.151 Migrant -0.196 0.452 -0.418** 0.151 Protest Timing Weekend 0.571* 0.259 0.027 0.087 Protest Form Static 0.347 0.265 -0.016 0.090 Protest Year 2004 -0.668 0.487 -0.176 0.162 (ref = 2003) 2005 0.049 0.473 -0.108 0.160 2006 -0.515 0.499 0.016 0.161 2007 -0.688 0.521 -0.198 0.166 2008 0.520 0.429 0.278 0.147 2009 -0.197 0.445 -0.231 0.148 2010 -0.296 0.469 -0.204 0.157 Constant -2.355*** 0.638 4.616*** 0.207 Alpha 0.568 0.035 Model Chi² 70.78*** 133.98*** Pseudo R² 0.097 0.024 Notes: * p <.05 ** p <.01 ***p <.001; df = 17; N= 461; size variable log transformed because of negatively skewed distribution. 96 Mobilization comfortable position to stage events that correspond to media routines. Moreover, the old social movements are primarily unions (51% of all OSM protests in the database are staged by unions, accounting for 27% of all protest events in the news) and professional organizations (staging 14% of all OSM protest, accounting for 7% of all protest events in the news). The protest events of unions and professional organizations—being the most resourceful and institutionalized social movement actors in neocorporatist Belgium—have especially high chances of media selection (26% and 24% respectively). A second, more issue-related argument might also be at play. First, the issues of old social movements are already far more prominent in the news. This favorable opportunity structure increases the odds of media exposure. Second, migrant organizations make claims about issues that are rarely considered relevant for the Belgian audience, which obviously raises media-selection thresholds (Wouters, De Swert, and Walgrave 2009). Previous accounts of the impact of the logic of disruption were mixed. In this case, how- ever, when protest events pose a threat and are disruptive, television coverage becomes more likely. Protest events with demonstrators behaving violently, getting arrested, holding up traffic or blocking entrances of buildings, more easily pass the media gates. Chances of media cover- age rise to 21 percent when demonstrators are disruptive. In order to be judged newsworthy, behaving in ways authorities perceive to be disruptive helps. In the same vein, demonstrations that need the deployment of at least one water cannon, or when a counter demonstration is ex- pected, are more likely to become news, albeit to a lesser extent (a rise of 5%). Hypothesis 6a is confirmed. Demonstrations that were not given permits, finally, do not make it more easily in the news: results for hypothesis 7 were not significant. What about the logic of bearing witness? Results show symbolic actions are significantly more newsworthy than actions without such elements. Symbolic actions have a 17% chance of making it into the news. Chances of media coverage rise 12% if demonstrators use symbolic action. In order to test whether the different logics of protest operate independently or affect each other, interaction terms were computed (not presented in the table). Whereas the logic of numbers and disruption (B = – 0.076; sig = 0.486) and disruption and bearing witness (B = 0.471; sig = 0.506) were not significantly related, the logic of numbers is negatively related to actions being symbolic or not (B = – 0.212; sig = 0.028). In other words, the positive effect of symbolism holds for small actions. For organizations that cannot or choose not to rely on the logic of numbers or the logic of disruption, staging symbolic actions thus clearly stands out as an alternative route towards media attention. Finally, weekend protests, make it into the news more easily (H12). Weekend protests appear to be somewhat larger, although controlling for demonstration size shows that it is slow news days that account for this higher rate of coverage. Across the different years under study, media selection appears to be quite stable. Only the year 2008 sees a significant increase in the number of protest events being covered.19 Next, models 2 and 3 allow us to assess differences in the filtering mechanisms between both television stations. By and large, the same input variables hold for the commercial and public broadcaster. The form of an event only loses significance (p = 0.066) for the public broadcaster, which covers more protest events. Of greater interest, however, are the differences in coefficients. Media ownership matters for the selection of protest events, and does so in the expected direction: events that do make it into the commercial newscast score higher on almost every determinant included in the model. Variables that measure violence and theatrics have much stronger predicted effects in the public broadcaster model. In brief, the commercial station predominantly selects those protest events that offer the drama factors our determinants measured, therefore generating the better fit of the model and the higher coefficients. Finally, models 4 and 5 further corroborate this conclusion. Focusing only on those events that were deemed newsworthy on one particular newscast, the models clearly show a more sensationalist and entertainment-oriented selection mechanism on the commercial channel (disruption and symbolic actions as highly significant predictors) and a more institutionally oriented (old social movements) media filter on the public broadcaster.  Mobilization Table 5. Coefficients of Logistic Regressions Predicting Any, (Unique) VRT and VTM Coverage of Protest Events in Brussels, 2003-2010. Model 1 Model2 Model 3 Model 4 Model 5 Any Media Coverage VRT Coverage VTM Coverage Unique VRT Unique VTM B S.E. B S.E. B S.E. B S.E. B S.E. Logic of Numbers Size 0.553*** 0.044 0.564*** 0.047 0.582*** 0.055 0.294*** 0.053 0.172* 0.084 Logic of Disruption Disruption 1.123*** 0.204 0.828*** 0.210 1.307*** 0.225 0.146 0.278 1.149** 0.355 Anticipated Disruption 0.674*** 0.147 0.635*** 0.153 0.702*** 0.182 0.358 0.195 0.418 0.326 Unexpected 0.317 0.247 0.410 0.259 0.329 0.332 0.350 0.312 -0.072 0.568 Logic of Bearing Witness Symbolic Action 1.298*** 0.174 0.957*** 0.190 1.453*** 0.217 0.666** 0.224 1.602*** 0.297 Sponsoring Org. OSM 1.071*** 0.219 1.096*** 0.236 0.958** 0.293 0.945** 0.284 0.548 0.456 (ref = other) NSM 0.758** 0.235 0.743** 0.256 0.809* 0.303 0.503 0.307 0.471 0.462 Migrant -0.368 0.223 -0.356 0.241 -0.445 0.303 -0.203* 0.289 -0.224 0.461 Protest Timing Weekend 0.626*** 0.137 0.562*** 0.146 0.685*** 0.178 0.403 0.177 0.578* 0.278 Protest Form Static -0.355* 0.145 -0.282 0.153 -0.518** 0.183 -0.015 0.203 -0.431 0.325 Protest Year 2004 0.213 0.277 0.368 0.292 0.297 0.352 0.087 0.355 -0.395 0.594 (ref = 2003) 2005 0.018 0.269 0.091 0.287 -0.023 0.351 0.111 0.338 -0.080 0.529 2006 0.108 0.266 0.282 0.283 0.240 0.343 0.030 0.345 -0.486 0.571 2007 0.037 0.274 0.103 0.294 -0.076 0.365 0.169 0.341 -0.063 0.531 2008 0.744** 0.247 0.843** 0.264 0.814* 0.317 0.551 0.314 0.221 0.480 2009 0.501 0.251 0.434 0.272 0.567 0.325 0.348 0.319 0.506 0.459 2010 0.067 0.266 0.292 0.281 0.041 0.344 0.125 0.337 -0.774 0.597 Constant -3.912*** 0.333 -4.237*** 0.360 -4.810*** 0.441 -4.193*** 0.428 -4.732*** 0.652 Model Chi² 820.38*** 698.04*** 632.90*** 156.10*** 101.11*** Pseudo R² 0.285 0.272 0.330 0.096 0.138 Notes: * p <.05 ** p <.01 ***p <.001; df = 17; due to missing values total N = 4,080, media N = 469. Size log transformed because of negatively skewed distribution. 98 Mobilization Media ownership matters for news selection, and the effect goes in the expected direction, con- firming hypothesis 2. Besides the mere selection of protest events, it is clear that coverage of protest can vary greatly. Here I test for two types of news item “packaging” (Wilkes, Corrigall-Brown, and Myers 2010). More precisely, I assess whether the determinants of media selection are related to the length of protest coverage and whether they affect the prominence of the event in the newscast.20 In other words, can these factors also predict a second step of media selection— volume and prominence of media attention? Results in table 5 show that even among those events that are singled out as newsworthy, large and disruptive demonstrations more easily make headline news and are discussed at greater length. The disruption finding positively con- firms the saying “if it bleeds, it leads.” Hypotheses 5b and 6b are confirmed. Who demonstrates seems to have no importance when it comes to making headlines. More professionalized organizations more easily pass media selection hurdles, but they do not make headlines more easily. This is somewhat different for the volume of coverage: coverage of migrant and asylum-seeker protest is significantly shorter than coverage of protest staged by every other organizer. Although migrant organizations and asylum seekers protest the most, they are least likely to make it into the news, and even when they make it onto the screen, attention will be short. The logic of bearing witness, finally, allows events to enter the mass media arena, yet symbolic events are not more likely to result in headlines or more lengthy coverage (H9). Differ- ences between both stations are small: for the public broadcaster, weekend demonstrations more easily make headlines. In the case of the commercial station, unexpected events became significant as headline predictors. Table 6 summarizes the multivariate findings for all hypotheses. Table 6. Summary Table of Hypotheses News News Likelihood of Television Presence Prominence Volume Routine Value Coverage H1 ; VRT > VTM + H2 VRT sensationalist < VTM ; + sensationalist H3a + 4a ; VRT < VTM - - H3b + 4b ; IF (VRT & VTM), VRT > VTM + + H5a + b ; Logic of numbers Ê + + + H6a + b Logic of disruption: (anticipated) ; + + + conflict Ê H7 Logic of disruption: ; ns ns ns unexpectedness Ì H8 ; Logic of bearing witness Ê + H9 ; Logic of bearing witness Ì ns ns H10 ; Static Ì + ns ns H11a ; ; OSM >NSM ns ns ns H11b ; ; NSM > Migrant + ns + H12 ; Weekends Ê + + ns CONCLUSION AND DISCUSSION This study scrutinized the selection of protest events by television news. Competition over media access is high and chances of media coverage are slim. I systematically assessed the impact of a wide host of theoretically relevant predictors on media selection, prominence, From the Street to the Screen 99 andvolume. Results based on data spanning eight years of protest show that only a few demonstrations make it from the street to the screen. About one demonstration in ten passes the media gates; the vast majority of protests are ignored. The public broadcaster, with its focus on hard news, covers many more protests than the commercial station. The latter has a more distinct and sensationalist selection profile. Media ownership matters for what is singled out as newsworthy. In line with previous findings for the press, television attention for demonstrations is far from arbitrary. Demonstration size appears to be the single strongest predictor of news- worthiness. Demonstrations that succeed in drawing many people to the street signal a relevant social issue journalists cannot ignore. Who demonstrates matters as well. Protests organized by more professionalized organizations—OSM in this case—more easily survive media selection hurdles. Yet, the fact that the issues of old social movements are featured more frequently in the news can also account for this result. Findings about the relevance of disruption in earlier newspaper research are mixed. Results for the television news in Brussels shows that disruption matters. Protest events that involve arrests, violence, or blockage of traffic are more likely to gain coverage. Disruptive demon- strations result in spectacular images. However, it is doubtful whether these outbursts of vio- lence and disruption generate public sympathy for the activists’ cause. Most protest actions, however, do not succeed in drawing many people to the streets, and violent and disruptive demonstrations are rare. Beside the logic of numbers and the logic of disruption, this study also assessed the logic of bearing witness. No previous studies have looked at protest events that follow this logic and their likelihood of becoming news. This lack of evidence is strange, as it is clear that symbolic events are keen adaptations of traditional protest forms to the operational logic of mass media. Results show that carefully staging and scripting drama makes television exposure more likely. This finding should encourage activists who cannot rely on numbers or do not want to act violently. They can increase their chances of securing television coverage by clearly staging and scripting protest events that offer compelling images and theatrical drama. New social movements especially have adopted this logic of bearing witness, making their presence in the media arena less a function of demonstration size. Whether the logic of bearing witness also works for activists seeking print coverage requires further investigation. Besides the selection of protest events, this study focused on the packaging of protest by television stations as well. Which factors determine prominence and length of news item(s) on protest? Both the logic of numbers and the presence of disruption account for higher profile (i.e., headlines) and lengthier coverage. Interestingly, when it comes to making news headlines it does not matter who demonstrates. This finding does not hold for the volume of coverage, how- ever. Protest organized by asylum seeker and migrant organizations are given significantly less airtime. Interestingly, the logic of bearing witness is a significant and strong factor for the selec- tion of events, but is insignificant in determining media volume and prominence. This finding provides room for speculation. Sobieraj (2010), interviewing journalists and activists at presi- dential conventions, concluded that media-friendly activist groups appear less appealing to the press because they are perceived by journalists to be less authentic. Whereas the results reported here counter the finding of Sobieraj when it comes to media selection, it could be that her conclusion is confirmed when it comes to media prominence and volume. Because of the ready- made character of symbolic actions, journalists feel tricked into covering these events, and want to restore the journalist-source power balance by denying these events headline status or lengthy coverage. Although these findings on media packaging certainly contribute to our understanding of the representation of protest groups in the mass media arena, this study does not engage more qualitative and content-related features of protest. The impact of protest is not only a function of being singled out as newsworthy, but also of how protest is filtered and presented through the journalistic lens. Does the spin of the news item favor the activist perspective? Or do media 100 Mobilization portrayals of protest marginalize and criminalize activists, their purposes and means? Both public and political reactions to protest are likely to depend on these descriptive characteristics of television coverage and therefore should be explored in future research. Nevertheless, further inquiry into media selection of protest events is also much needed. This study started from a very static perspective of newsworthiness. It looked at characteristics of protest events to determine likeliness of television coverage. Newsworthiness is not merely the product of objective characteristics of an event, however. Events also make it into the news because of their relation to other events. Newsworthiness then is not so much an objective quality of a particular event, but the interpretation by journalists of event characteristics in the light of other events. In other words, the outcome of journalistic gatekeeping activity is influ- enced by what already made it into the news. Social movement and media scholars have there- fore emphasized the dynamics of media and issue attention cycles, a term coined by Anthony Downs (1972). Issue dynamics are rarely empirically tested, however, and the degree to which different kinds of organizations depend on a favorable and suddenly sympathetic media environ- ment remains by and large unexplored. Future research should therefore do more to incorporate the dynamics of past media attention when trying to explain likelihood of media coverage. Such research will acknowledge to a greater degree the importance of timing for protestors working to diffuse their claims, just as it would allow researchers to explore whether and when protest actions ride the waves of previous media attention, or instead cause the swell and set the agenda. Only by proceeding along these lines of research and conducting rigorous empirical tests, news making will become the substantive building block of social movement theory Kielbowicz and Scherer (1986) desired it to become more than twenty-five years ago. NOTES 1 Generally, a news carrousel consists of a number of very short news items that quickly follow each other, accom- panied by up-tempo background music and a continuous voice over. 2 Though I write Belgium, I actually mean Flanders, the northern part of the country. It contains 60% of the population. 3 The term “newshole” is used to indicate the amount of space or time in a paper or a newscast. 4 The fact that scope matters can be the consequence of two mechanisms. First, a news routine mechanism can be at work: journalists can more easily report nearby events. The reason for selection then is purely a matter of logistics. Second, a news value mechanism might play as well: audiences are less interested in distant news, as it is considered to have a less powerful impact on their personal life. 5 Each newscast on both stations lasts approximately half an hour, and deals with the same amount of items. For the entire 2003-2010 period, a typical VRT (public broadcaster) newscast lasts about 2,164 seconds (36 minutes) and shows 23 news items. A VTM (private station) newscast lasts about 2,166 seconds (36 minutes) and shows 22 news items. In other words, the stations are extremely alike when these formal features are taken into account. 6 The fact that symbolic actions clearly adapt to the logic of mass media does not mean that the sole purpose of these events is attracting media attention. Just like any other type of protest activity, participation in symbolic actions can be a bonding activity, reinforcing collective identity and boosting commitment. Because of the explicitly scripted nature of the action, which suggests media attention is a core objective, I regard these effects on participants more as byproducts, however important they might be for the sustainment of the movement. 7 The location of the protest is also crucial in this respect: static protests at symbolic locations—in front of embassies or parliaments, for example—are obviously more directly persuasive and specifically targeted than static protest on a symbolic yet large square. 8 I would like to thank Commissioner Hannelore Hochepied and Head of the Archive Walter Reniers for invaluable support. 9 The share of After-Demonstration Reports—or so-called Post-RARs (Rapport Administrative / Administratief Rappport)—was not evenly distributed across years. For the period 2006-2010 high degrees of Post-RARs were incorporated (on average 70%). For 2005, the share of Post-RARs was only 20%. The year 2004 contained 52% Post- RARs, 2003 had a high level of after demonstration information (67.3%). Regression results did not change when only those demonstrations with Post-RAR were included. 10 This is not necessarily problematic as journalists only know the estimates in advance when they decide to cover the protest. Moreover, if protest organizers fail miserably and a disappointing turnout is the mobilization outcome, this will also be news. Using estimates is only problematic when a protest was expected to draw a small number of participants but instead attracted many. Yet as mobilization (at the individual level, at least) is thought of as a process involving different stages with potential participants dropping out from one stage to the next, this latter situation can be thought of as least frequently occurring.  From the Street to the Screen 101 11 In total, 82 protest actions were reported but not incorporated in the police records, an average of 10 actions a year, or 14 percent of all demonstrations with media coverage. A look at these missing demonstrations does not show a systematic bias: big as well as small events, organized by all kinds of groups are covered by mass media, but were not found in the police archive. 12 For the 1999-2009 period, digital data (with limited information) were made available about all the records in the police archive (including the nonprotest records). In this period, 35,637 records were created in the archive. On average 15.5% of all police records mentioned synonyms of “protest” or “protestor.” Interestingly, the share of protest records in the database increased over time. 13 Data from the study centers of both the VRT (public broadcaster) and the VMMA (commercial group) reported in Wouters, De Swert, and Walgrave, 2009, “A window on the world,” Flemish Peace Institute report. 14 The file that performs the needle haystack word search and creates the protest dummy variable in the news media database is available upon request. The syntax was run on both a full text and a keyword summary variable of the database. The police database and news media database were manually linked. I would like to thank Dr. Jonas Lefevere for his help with the construction of the syntax. 15 Oliver and Myers (1999) and Oliver and Maney (2000) used two different operationalizations of media coverage. In the 1999 article they focused on “timely coverage” as references to an event in an interval of 31 days around the event. In the article published in 2000, they use a reference in the 12-month period around the day of action as a measure of media coverage. The authors conclude that many events get covered independent from their day of action (25%). This characteristic seems to be idiosyncratic to (local) newspaper coverage. In this study, prior media references to future protest events in the television database were found only when unions announced a general day of strike, with focus on potential traffic jams and other discomfort. Post demonstration coverage was limited to reports about damage, arrests, or court decisions about the latter. Generally, televised coverage of protest is a same- day attribute, completely in line with the more fast and current affairs focus of television. 16 Note that both the public and commercial station have about the same amount of “headline” items every newscast (4.7 on average for the commercial station, that is, about 22 percent of all news items within a typical newscast; 5.7 on average for the public broadcaster, that is, about 26 percent of all news items). 17 VRT (Min = 15; Max = 1,192; St.dv = 106; 5% trimmed mean = 93); VTM (Min = 18; Max = 775; St. dev. = 96; 5% trimmed mean = 95). 18 Average number of expected participants over all demonstrations is 437 participants (5% trimmed mean is 129 participants). Nonsymbolic Actions mean = 444 participants / 5% trimmed mean = 133 participants. Symbolic Actions mean = 360 participants / 5% trimmed mean = 94 participants 19 Although it is difficult to relate this increase in amount of coverage to a single cause, I speculate that the start of the economic crisis (more protest, and, protest could be another angle for journalists to cover the crisis) together with a lack of a federal government (resulting in less official news), may have been potential grounds for higher media selection of protest. 20 Both prominence and volume are strongly correlated; that is: prominent items are also more voluminous. Logically, news headlines are treated more in length (Pearson r = 0.673; Sig. = 0.000). Appendix A: Correlation Coefficients 1 2 3 4 5 6 7 8 9 10 11 1 Size 1 2 Conflict.159* 1 3 Anticipated.272*.208* 1 Conflict 4 Unexpected -.132*.112*.318* 1 5 Symbolic -.080* -0.19 -.100* -,018 1 Action 6 Old Social.238*.088*.049* ,022 -,027 1 Movement 7 New Social -.193* -0.27 -.068* -,021 ,293** -,253* 1 Movement 8 Migrant/.014 -.061* -.019 -,019 -,195** -,492* -,419* 1 Asylum Org 9 Other -.104*.011.042* ,024 -,010 -,224* -,191* -,372* 1 Organization 10 Weekend.099*.020 -.095* -,061* ,009 -,095* ,052* -,037* ,109* 1 11 Static -.479* -.145* -.216 ,055* ,017 -,216* ,061* ,118* ,026 -.125* 1 Note: * = p < 0,01