Less than Expected? How Media Cover Demonstration Turnout

Abstract

Demonstration turnout is a crucial political resource for social movements. In this article, we investigate how mass media cover demonstration size. We develop a typology of turnout coverage and scrutinize the factors that drive turnout coverage. In addition, we test whether media coverage underestimates, reflects, or exaggerates “guesstimates” by organizers and police forces. Together, these analyses shed light on whether turnout coverage fits a logic of normalization or marginalization. We rely on a unique dataset of 428 demonstrations organized in Brussels (2003–2010). For these demonstrations, we have information on the turnout as reported in national television news, as counted by the police, and as expected by the organizers. We find that media present turnout most often as a fact, rarely as contentious (10 percent). Although few demonstrations pass the media gates, our study yields little to no evidence for a logic of turnout marginalization. Media coverage does not systematically underestimate demonstration size, nor does it blindly follow police counts. Rather, turnout coverage attests of a logic of normalization, following standard news-making practices. The more important the demonstration (size, lead item) and the larger the gap between police and organizer guesstimates, the more attention is paid to turnout in the news. Discussion centers on the generalizability and normative interpretation of the results.

Introduction

One of the most important aspects of protest politics is the size of the crowd an organizer succeeds to mobilize. In representative democracies, “power is in numbers” (De Nardo 1985). Demonstration turnout signals the significance of a movement, testifies of its strength, and is a critical resource for social movement bargaining power (Burstein 1999; della Porta and Diani 1999; Lohmann 1993). Large crowds not only attest of successful mobilization, they also indicate a demonstration’s likely success in the media and political arena (McAdam and Su 2002; Tilly 2004; Walgrave and Vliegenthart 2012). Given the importance of protest size for demonstration outcomes and given that most observers learn about demonstrations indirectly via mass media (Cottle 2008; Ferree et al. 2002; Koopmans 2004), this article tackles a straightforward question: How do mass media report demonstration size?

Questions of how media portray demonstrations have attracted the interest of both social movement and communication scholars (Boyle et al. 2004; Gamson 2004; Smith et al. 2001; Weaver and Scacco 2013; Wouters 2015). The standing conclusion in both fields is that media coverage tends to undermine social movement agendas. Communication scholars speak of a “protest paradigm,” a fixed template that journalists use to cover protest (Mcleod and Hertog 1998). Elements of this script are a focus on (incidental) details of the protest rather than on the substantial issue the protesters raise, the use of official rather than protester sources to define the problem at hand, and the use of news frames that present the protestors as minority extremists rather than informed and rational “good” citizens (Boyle et al. 2005; Hertog and McLeod 1995; McCluskey et al. 2008). In some formulations of the protest paradigm, media organizations are considered to be lapdogs of those in power, protecting the status quo by “marginalizing, criminalizing and demonizing” protesters (Entman and Rojecki 1993; McLeod and Detenber 1999; Olien et al. 1989).

More recent research, however, has made a beginning with exploring the conditions under which media tend to marginalize protest. These studies explore variation in adherence to the paradigm and even ask whether and to what extent media coverage might actually be advantageous to social movements (Amenta et al. 2012; Evans 2016; Lee 2014; Shahin et al. 2016; Taylor and Gunby 2016; Wouters 2015). One mechanism that could explain why media do not blindly follow the paradigm lies in the trend of protest normalization (Van Aelst and Walgrave 2001). In the past few decades, protest—especially in the form of street demonstrations—has become a more legitimate means of signaling grievances in Western democracies, albeit more so in Europe than in the United States (Caren et al. 2011; Dalton 2008; Klandermans 2012; Meyer and Tarrow 1998). With an increasing number of people from more diverse segments of society participating in demonstrations, it seems plausible that audience-dependent mass media are less apt to systematically marginalize or undermine protest. Instead of being based on the protest paradigm script of marginalization, coverage then is likely to be based on more standard news-making practices guided by professional norms and principles of newsworthiness (Bennett 1990; Shoemaker and Reese 1996).

To explore media representation of demonstration turnout, we develop a typology of turnout coverage and scrutinize the factors that drive turnout coverage types. Next, we compare protest size as mentioned in television news with “guesstimates”¹ by police and protest organizers. Together, these analyses allow us to assess whether reporting protest size is more a matter of marginalization (media, by intention or not, undermining social movement agendas by distorting the numerical strength of protest) or rather a matter of normalization (journalists following standard news-making practices when covering demonstration turnout). Surprisingly, little systematic effort has been made to analyze how journalists report protest size (for a notable exception, see Mann 1974). The main reason for this scholarly neglect, we believe, is that the concept of demonstration turnout is complex and systematic data on turnout are rare. Indeed, although turnout comes across as a self-evident truth, it hardly ever is an indisputable fact. Views on demonstration size are most often the consequence of guesstimates made by journalists, organizers, and police department(s). Each of these actors has its own interest and take on the issue, frequently leading to diverging counts and sharp debates (Biggs 2016; McPhail and McCarthy 2004; Van Aelst and Walgrave 1999). Despite the interest of many media scholars in aspects of “fairness,” “balance,” “bias,” and “objectivity” (Danielian and Page 1994; De Swert 2011; Entman 2007; Hopmann et al. 2011), the elusive nature of demonstration turnout and the lack of systematic data might have discouraged up-close scrutiny.

In this article, we use a unique dataset of 428 demonstrations organized between 2003 and 2010 in Brussels, the capital of Belgium. Brussels hosts many national and international political institutions and is a demonstration hotspot. Our dataset contains all demonstrations organized in Brussels that attracted television news attention during that period. For these demonstrations, we have systematic information on the turnout as reported in television news, as counted by the responsible police department, and as expected by the organizers before the demonstration. Although we acknowledge that demonstration turnout cannot be exactly measured, we believe that comparing these guesstimates—which are very likely the sources journalists turn to when constructing news—can help us to better understand processes of bias, power, and control in the media arena (Carragee and Roefs 2004). The specific media arena we focus on is the Belgian television news landscape. With two prime-time newscast competitors (one commercial and one public) being true broadcasters, this media environment strongly resembles many other European media landscapes, yet is notoriously different compared with the U.S. media environment typically scrutinized in protest paradigm studies. Our results show that Belgian television news presents turnout most often as a fact, devoid of its contentious nature. We find little to no evidence, however, of media systematically underreporting organizer’s expectations, nor of media blindly following police figures or treating some organizations or issues systematically differently. In all, the patterns we find do not fit the dominant perspective of a hostile media, protecting the status quo by marginalizing protest and downplaying numerical strength. Rather, the mechanisms that underlie our findings suggest the existence of rational, professionalized journalists guided by standard news-making practices.

A Turnout Typology

How do journalists report demonstration size? We distinguish two main types of turnout coverage. In case of factual turnout presentation, a single, more or less exact indication of protest size is given in a news report. Contentious turnout presentation, on the contrary, comes in two guises: one of interpretation and one of contestation. “Interpretation” refers to the use of an external baseline that allows for the evaluation of the turnout. For instance, the turnout can be more or less than expected, or, the turnout can be lower or higher compared with a similar demonstration in the (recent) past. The second type of contentious turnout coverage is “contestation.” Here, multiple and diverging counts are given within a single news report, clearly indicating that the turnout is contested. For instance, the number counted by the police could be lower (or higher) compared with the turnout according to the organizers. Figure 1 presents the typology.

Figure 1.

A typology of turnout coverage.

Our typology resonates with much previous literature on how media present information and as such, challenge or re-create power relations (Carragee and Roefs 2004). Specifically, the distinction between factual and contentious coverage echoes ideas of Hallin (1986) and Gamson et al. (1992) on distinct spheres of media discourse. One sphere is uncontested (factual). Media content in this sphere is presented as a transparent description of reality. What is reported in this realm is presented as taken for granted, common sense, and in a way, depoliticized. The other realm is of a more contentious nature. Here, media coverage explicitly presents cues to the audience about a struggle over meaning and power. The coverage contains multiple views on an issue, offers interpretation, and as such, is far less of an (apparently) value free nature. Applied to our case, coverage of demonstration turnout can fall in the uncontested realm, with (a single) turnout (figure) presented as a dry fact, or in the contentious realm, showcasing conflict and interpretation.

The first goal of this article is simply to map the extent to which turnout is presented in either form in Belgian television news. In addition, however, we aim to understand the factors that drive turnout coverage. Understanding these drivers might help us to better understand whether a logic of marginalization or normalization is at play in turnout presentation. Based on protest paradigm literature, certain issues and organizations might trigger the protest paradigm given their threat toward the status quo. If the type of turnout coverage is driven by organizational or issue characteristics of the protest, then clearly not all organizations or issues are treated equally in the media arena. This might signal that elements of a logic of marginalization are at work in the media sphere. That is, protest, depending on the kind of issue it tackles or the kind of claim it puts forward, is treated differently compared with protest that tackles other, perhaps less controversial issues. For instance, turnout of union protests might be treated differently given their status in neo corporatist Belgian politics, or certain new social movement issues, given their identity politics claims, might be treated differently compared with bread and butter issues. Boyle et al. (2004) already showed that issues matter in protest coverage, as they affected the framing and tone of protest in newspaper articles (also see Boyle et al. 2012). Hence, if turnout coverage turns out to be a function of the issue and the type of staging organization, then clearly not all issues and organizations are treated equally, adding evidence to studies who claim that a logic of marginalization is at work in the media arena.

However, if journalists approach turnout in a more routine manner, following news values and professional standards and norms, then turnout coverage would better fit a logic of normalization. In this scenario, two event size characteristics, in line with news value theory (Galtung and Ruge 1965; Harcup and O’Neill 2001; Shoemaker and Reese 1996), are especially likely to affect how turnout is covered. First, one could expect that more significant demonstrations—demonstrations that attract many participants—are more likely to receive contentious turnout coverage. Given the exceptionality and potential consequences of large demonstrations—such demonstrations are more likely to alter elements of the existing social order—journalists are more likely to pay attention in their report to the magnitude of the crowd, either by evaluating the significance of the turnout (interpretation) or by presenting the audience with conflicting turnouts (contestation). Second, also when protest organizers and police departments strongly disagree about the number of protest attendants, the odds of more attention to turnout (and hence contentious turnout coverage) are likely to rise. In cases of conflicting counts, journalism textbooks would advise the professional journalist to cover this newsworthy element, not only because it signifies conflict and makes an interesting story, but also because it attests to the norm of fair and balanced coverage. Therefore, if event characteristics turn out to drive turnout coverage, a logic of normalization might be a more appropriate interpretation of how journalists present protest size.

In all, this leads to two contrasting hypotheses. Turnout coverage could either be a function of issue and organizational characteristics (testifying of a logic of marginalization), or turnout coverage could be a function of event characteristics related to the magnitude of the crowd itself (testifying of a logic of normalization).

• Hypothesis 1 (H1): News coverage of demonstration turnout is a function of issue and organization characteristics. (marginalization hypothesis)

• Hypothesis 2 (H2): News coverage of demonstration turnout is a function of event size characteristics. (normalization hypothesis)

• Hypothesis 2a (H2a): News coverage of demonstration turnout is more likely to be contentious when the demonstration is of higher significance, as measured by demonstration size.

• Hypothesis 2b (H2b): News coverage of demonstration turnout is more likely to be contentious when the gap between police count and organization expectation increase.

Comparing Guesstimates

Next, also comparing media turnout to “guesstimates” made by organizers and police might improve our understanding of how media report turnout. Media can either underestimate, accurately reflect, or exaggerate police and organizer guesstimates. Or, combining both guesstimates, journalists can also search for a middle ground and present the average.

Most of the extant references on media and demonstration size focus on how media tend to underreport or discredit turnout. The seminal work of Gitlin (1980) speaks of “disparagement by numbers” as a framing device media use to denigrate movements. A number of examples are presented throughout his book. For instance, coverage of the March on Washington (April 17, 1965) in the New York Times claimed “more than 15,000 students and a handful of adults” to be present, whereas police estimates and coverage of the (left-wing) weekly The National Guardian spoke of 25,000 protest attendants (Gitlin 1980: 49–60). Gitlin (1980: 95) gives another example of how 1,000 pro-war demonstrators received as much coverage as 10,000 antiwar demonstrators, arguing that this distribution of attention marginalizes some protest and legitimates others. Similarly, Parenti (1986), in his book “Inventing Reality,” presents “typical examples of how the press treats protests on the Left.” A Washington Post report, covering the “March on the Pentagon” (May 3, 1981), for instance, only presented the police estimate of 25,000 participants and made no mention of the 100,000 participants claimed to be present by the organizers. Or, the turnout of the 1981 New York labor day parade—mobilizing 200,000 according to the organizers and 100,000 according to the police—is presented as “a disappointingly small crowd of less than 100,000 union workers.” The examples given by the authors in this paragraph lead to a clear conclusion: Media tend to downplay a movement’s numerical strength, attesting to a logic of marginalization (for similar examples, see Brasted 2005; Entman and Rojecki 1993).

McCarthy et al. (1998) present less exemplary evidence on coverage of demonstration turnout. Comparing the number of participants as expected by the organizers (stored in the archives of Washington, D.C. police) with the turnout as presented in U.S. quality newspapers, they conclude that newspapers do “a reasonably good job of reporting hard news items of protest demonstrations correctly.” Specifically, the correlation (p < .001) between permit and newspaper turnout reached up to .872, indicating a strong link between both measures. Although correlation does not directly tap undercounting—in fact, media might systematically undercount organizer’s estimates and still yield high correlations—the contribution of McCarthy et al. is valuable as the authors suggest that media might actually quite fairly reflect demonstration size, attesting to a logic of normalization.

In sum, again, literature is not on the same page. How can we make sense of tilt in turnout coverage? Our typology presents a useful starting point. In case of contested turnout coverage, different guesstimates and the resulting conflict over numbers is made quite explicit in the news. In case of factual presentation, however, this struggle over meaning is obscured for the viewer. The journalist presents a single count and leaves the audience c(l)ueless. Therefore, especially in case of factual turnout presentation, the relationship between guesstimates and turnout in the news seems relevant. If media systematically underreport organizer expectations or consistently follow the count of the police and ignore the perspective of the organizer, factual turnout coverage might not be so innocent. Rather, hidden from the public, a pervasive logic of marginalization then would underlie what journalists present as “facts.” The absence of such a pattern, however, might again be an indication that media coverage of protest size is more a matter of normalization.

To address this line of thinking, we will first explore the relationship between factual turnout coverage, police, and organizer guesstimates. Next, we will investigate factors that drive undercounting, reflecting, or exaggerating. Again, we hold (1) that issues and organizations, (2) the size of the demonstration, and (3) the gap between police and organizer guesstimate might matter for tilt in factual turnout presentation. First, if protest turnout on some issues or by some organizations is more likely to be distorted compared with protest on other issues, then clearly not all issues are treated equally and one could reason that there is some bias at play in the media arena. Second, event size might matter. According to Gans (1979), the higher the turnout of protests, the more of a threat protesters present to the existing social order, and the more likely attendance to these protests would be marginalized. If especially participants at large demonstrations are undercounted, this too might add to a logic of marginalization. Third, if journalists systematically prioritize police counts when reporting demonstration size, their dependence on this official source, according to critical theorists, would testify of status quo bias, easily reflecting and strengthening existing power structures. Wrapped up, confirmation of these hypotheses would present evidence that the reported number of participants in factual turnout coverage of protest is driven by a logic of marginalization. If the hypotheses are rejected by the data, however, a normalization-interpretation becomes more likely.

• Hypothesis 3 (H3): Undercounting the number of participants in the news is a function of issue and organization characteristics. (marginalization)

• Hypothesis 4 (H4): The larger a demonstration, the more likely the number of participants is undercounted in the news. (marginalization)

• Hypothesis 5 (H5): The larger the gap between organizer and police guesstimate, the closer the number of reported participants sticks to the police count. (marginalization)

Data and Methods

This paper combines two data sources. Dataset 1 was manually collected and coded from the archive of the police district Brussel-Hoofdstad-Elsene. Brussels is the political capital of Europe and Belgium. With its many political institutions (Embassies, European, federal and regional parliaments and cabinets), Brussels is a demonstration hotspot. The police archive contains protest permit applications submitted by organizers and keeps a separate file for each protest request. Each file consists of three elements: (1) the actual protest request (a letter by the organizers containing standard information, including the name of the organization, its claim, and the number of participants the organizer expects), (2) the decision by the police to grant permission or not, and (3) a report of the actual protest (describing how many participants showed up according to the police and to what extent the police had to intervene). We do not treat either of the turnout measures in the archive as exact counts of “reality.” Obviously, organizer expectations do not necessarily materialize and police estimates might be biased. Rather, we consider these “guesstimates” to be likely cues that might have influenced journalists. Between 2003 and 2010, 4,582 protest events are listed in the archive.

Dataset 2 contains content analysis data of those demonstrations that succeeded to attract television news attention. In total, 428 unique protest events were covered in the 19 o’clock flagship newscasts of the main public (Eén) and commercial (vtm) Belgian television station, resulting in 564 news reports (as some demonstrations appeared on both stations). The Belgian television news market is a textbook example of a duopoly situation: Only two stations provide prime-time newscasts that are characterized by strong convergence and large viewer shares and ratings (Hooghe et al. 2007). The number of television news items shows that attracting media coverage is very much an uphill struggle for protestors: Few demonstrations succeed to elicit media attention, most are ignored. Especially demonstrations that attract many attendants, are disruptive, offer symbolic drama, are organized by strong sponsors, and are staged at the right time in an issue-attention cycle are more likely to become news (Earl et al. 2004; Oritz et al. 2005). For a media selection analysis on these data, see Wouters (2013). Coding of the protest reports was done by one of the authors and three trained master students. With regard to demonstration turnout, the following elements were coded: (1) whether a turnout of the demonstration was mentioned, (2) whether a specific turnout figure was mentioned, and (3) whether multiple turnouts were mentioned. For each turnout that was mentioned, the following elements were coded: (1) who mentioned the turnout/to whom the turnout was ascribed (the journalist, the organizer, the police, another source), (2) a description of the turnout itself, and (3) whether an evaluation or interpretation of the turnout was made (less than expected, no evaluation, more than expected).

Our study scrutinizes turnout coverage in these 564 news items, associating data from the police archive dataset with the media content dataset. For 489 (87 percent) news items, we have a turnout expectation by the protest organizer as stated in the permit request; for 413 news items (73%), we have a count by the police department as stated in the police report; for 357 (63%) news items, we have both. For 275 news items (49%), we have full turnout information, that is, organizer, police, and media provide a turnout. For more methodological details about the police archive data collection and the media selection analysis, see Wouters (2013). For more methodological details about the media content analysis, see Wouters (2015). Table 1 gives an overview of the variables used in this study and their operationalization.

Table 1.

Overview of Key Variables.

Dataset	Variable Name	Operationalization
Media	Turnout mentioned	(0/1) The news item reports a turnout measure
	Factual turnout	(0/1 + turnout) The news item reports a single turnout measure
	Contentious turnout	(0/1) The news items reports turnout as interpretation, contestation, or a combination of both.
	. . . Interpretation	(0/1 + turnout) turnout is interpreted by means of a baseline
	. . . Contestation	(0/1 + multiple turnouts + sources) multiple turnout counts are presented
	Lead item	(0/1) the news item is the opening item of the newscast
	Public broadcaster	(0/1) the news item is aired by the public broadcaster
Police	Organizer guesstimate	(numeric) expected demonstration size by organizer
	Police guesstimate	(numeric) counted demonstration size by police
	Guesstimate gap	(categoric; −5 to 5) distance between organizer and police guesstimate, measured as the proportion of the organizer’s guesstimate that was counted by the police, running from more than 80% less turnout; to more than 80% more turnout, with 20% intervals
	Abs guesstimate gap	(categoric; 0 to 5) absolute distance between organizer and police guesstimate, measured as the absolute proportion of the organizer’s guesstimate that was counted by the police, running from equal guesstimates to more than 80% less or more turnout counted by the police, with 20% turnout intervals.
	Issue categories	(0/1) based on the claim of the organizers in the protest request, different issues were distinguished. We compare the following new social movement issues with all other issues: Human rights, Asylum, Peace, Environment.
	Organization type	(0/1) based on the name of the organization in the protest request, organizational types were distinguished. We compare trade union organization vs. other organizations.
	Large demonstration	(0/1) number of participants >1,000; as counted by the police. If no police guesstimate was available, organizer guesstimate was used to complete this variable

Results

Scrutinizing Media Coverage

How do journalists report demonstration turnout? Table 2 presents basic descriptives.

Table 2.

Basic Descriptives.

		n	%	Total
Turnout	mentioned?	414	73.4	564
	. . . by journalist	411	99.3	414
	. . . by organizer	17	4.1	414
	. . . by police	16	3.9	414
Type	Factual	372	89.9	414
	Contentious	42	10.1	414
	. . . Interpretation	30	7.2	414
	. . . Contestation	20	4.8	414

Table 2 leads to several interesting observations. First of all, mentioning demonstration turnout appears as an almost indispensable aspect of protest coverage: Three protest reports in four mention demonstration turnout (73.4 percent). Second, not all players voice their take on protest size equally frequently. Journalists have a quasi-monopoly position on mentioning turnout, presenting demonstration size in 99 percent of the turnout reporting news reports. Compared with journalists, organizer (4.1 percent) and police (3.9 percent) perspectives are clearly sideshows. Interestingly, organizer and police perspective on demonstration size tend to appear together (r = .719, p < .01), suggesting that journalists aim to balance their reports with a countervoice once one voice gains prominence. Third, and perhaps most striking, demonstration size is rarely contested (4.8 percent) or interpreted (7.2 percent) in the news. Most often, turnout is presented like a fact (89.9 percent).

Other descriptive patterns emerge, although in general the number of stories that provide contentious coverage of turnout is low. For instance, if organizers voice their perspective on turnout by means of a quote (n = 7), they always argue that more people showed up than expected. And, if both organizers and police voice their turnout estimate (n = 12), the estimate of the police is always lower compared with that of the organizers. Finally, when journalists evaluate demonstration turnout (n = 27), their interpretation swings both ways (higher than expected: n = 13; lower than expected: n = 14). All told, these descriptives suggest that the struggle over demonstration turnout in the media arena is far less fierce than what one would expect from extant literature. In the absolute lion’s share of turnout coverage, the contentious nature of turnout stays under the radar.

These results makes one wonder about the conditions that determine how turnout is presented. Table 3 presents results from several stepwise logistic regressions predicting contentious turnout coverage. We subsequently added organization and issue features (Model 1), event size characteristics (Model 2), and finally controls (Model 3) to the model, as such testing our first block of competing hypotheses. Statistically significant odds ratio’s are in bold (p < 0.05). H1 expected turnout coverage to be a function of issue and organization characteristics. H1 gets little to no support from the data; the explanatory power of Model 1 is low. The turnout of protest staged by more established organizations (unions) is not treated differently in the media arena compared with the turnout of less established organizations. And, similarly, no differences between protest issues are found. Peace demonstrations do receive more contentious turnout coverage, yet the effect is only marginally significant and disappears in later stages of the model. Most peace protests in the database were staged against the Iraq War initiated in 2003. In Belgium, as in many other countries in the world, both government and opposition opposed the Iraq war, backed up by a large majority of public opinion (Walgrave and Verhulst 2009). The contentious turnout coverage of Iraq war protests hence signals a nonroutine treatment of the issue; yet in this case, media “sided” by the protesters. Turnout coverage confirmed the significance of the movement; journalists, organizers, and police stressed the magnitude of the demonstrations. Only in later years, when the annual “birthday” of the war was commemorated and smaller crowds were mobilized, contentious turnout coverage started questioning the viability of the movement. In all, organizations and issues are of little potency in predicting turnout coverage, questioning whether a logic of marginalization is at play in the media arena.

Table 3.

Logistic Regressions Predicting Contentious Turnout (Models 1–3).

	Model 1		Model 2		Model 3
	Odds	Sig.	Odds	Sig.	Odds	Sig.
Marginalization
Union	0.952	.906	0.976	.966	0.971	.960
Human rights issue	0.781	.651	0.437	.326	0.407	.311
Asylum issue	0.148	.068	0.393	.398	0.361	.370
Peace issue	2.599	.054	3.131	.061	2.504	.149
Environmental issue	0.620	.677	1.588	.693	1.022	.987
Normalization
Large turnout			3.825	.008	3.865	.009
Abs. Guesstimate Gap			1.644	.003	1.578	.008
Controls
Lead item					4.220	.005
Public broadcaster					0.451	.078
Constant	0.122	.000	0.009	.000	0.028	.000
Nagelkerke R2	.060		.244		.308
N	414		275a		275

^aThe number of observations drops due to missing values on the Abs. Guesstimate Gap variable, as explained in the “Data and Methods” section. If we run Models 2 and 3 without that variable, the number of observations is constant across the models and explained variances reach 0.162 (Model 2) and 0.258 (Model 3).

Note. Bold signifies values below .05.

H2 expected turnout coverage to be a function of event size characteristics. H2 gets strong confirmation. The explanatory power of the model quadruples by adding two size-related variables. Especially larger demonstrations increase the odds of contentious presentation. Demonstrations with a thousand participants or more are about four times more likely to receive contentious turnout coverage (H2a). And, when the gap between the expectation of the organizer and the estimate by the police increases, odds of contentious turnout coverage increase as well (H2b). With every 20 percent increase in the organizer–police turnout gap, the odds of contentious turnout coverage increase with about 60 percent. In Model 3, finally, we add lead item status and broadcaster type as controls to the models. Headline status, in part a function of demonstration size itself (McCarthy et al. 1996; Wouters 2013), strongly increases the odds of contentious turnout coverage; broadcaster type has no effect.

In sum, the evidence in Table 2 strongly suggests that not so much issue or organization characteristics, but rather event size characteristics determine how turnout is presented. The more important the demonstration, the more interpretation the number of participants gets. And, the higher the gap between police and organizer guesstimates, the more likely the journalist will have both sides have their say. Turnout coverage hence seems to relate more to mundane processes of professional news making, adding to a logic of normalization.

Comparing Media Turnout with Police and Organizer Data

The analyses above tell only part of the story, however. Contentious turnout coverage might be guided by features that suggest routine, professionalized journalism, yet in 90 percent of the news reports, still a single turnout measure is reported. Which cues do journalists follow in these (many) situations? In the following paragraphs, we compare demonstration size as reported in the media (factual turnout coverage) with the estimates of police and the expectations of organizers. Doing so allows us to track the possible “tilt” in media portrayal that lingers under the surface of a news report. If journalists present only one specific figure, concealing potential debate about turnout, what kind of source do they tend to follow?

Table 4 compares factual turnout coverage with police and organizer guesstimates. In the rows, we compare the turnout as expected by the organizers to the turnout as estimated by the police. If the turnout expectation by the organizer fell in an interval between minus and plus 15 percent of the police turnout count, both guesstimates were considered “similar.” In the columns of Table 4, we compare police with media turnout applying the same logic. Together, rows and columns in Table 4 allow us to substantially analyze how journalists report demonstration turnout under three different conditions: when the expected turnout by the organizer was (1) lower, (2) equal, or (3) higher than the estimate by the police. Only news items with a reported turnout of at least one thousand participants are included in Table 4 to get a clean comparison.² What do we learn from comparing these guesstimates?

Table 4.

Comparing Organizer, Police, and Media Turnout Estimates.

	Media < Police Estimate	Media = Police Estimate	Media > Police Estimate	Total
Org expectation < Police estimation
Count	9	25	16	50
% row	18.0%	50.0%	32.0%	100%
% column	60.0%	53.2%	36.4%	47.2%
Org expectation = Police estimation
Count	3	12	8	23
% row	13.0%	52.2%	34.8%	100%
% column	20.0%	25.5%	18.2%	21.7%
Org expectation > Police estimation
Count	3	10	20	33
% row	9.1%	30.3%	60.6%	100%
% column	20.0%	21.3%	45.5%	31.1%
Total
Count	15	47	43	106
% row	14.2%	44.3%	41.5%	100%
% column	100%	100%	100%	100%

Eyeballing the last row of Table 4 shows that media rarely report a turnout that is smaller than the count of the police (14.2 percent). Journalists tend to follow the police count (44.3 percent) or report a figure that is at least 15 percent higher (41.5 percent). Breaking down these numbers across constellations of police versus organization estimates sheds further light on the issue of which cues journalists follow. The first row of Table 4 presents demonstrations that mobilized more people, according to the police, than the organizers initially expected. If mass media would be adversarial toward these protests, they would cripple protest power by reporting a turnout lower than counted by the police, closer to the expectation of the organizers. Our analysis shows that this is least frequently the case (18.0 percent). Rather, journalists follow the (higher than expected) turnout as counted by the police (50.0 percent), or report a turnout that is even larger than the police turnout (32.0 percent).

Row 2 presents information on events for which organizer and police guesstimates were about equal. More than half (52.2 percent) of the news items of these events indeed report a turnout similar to the estimate of both police and organizers. In a third of these events (34.8 percent), media actually report higher turnouts than police counts. And, again, least frequently, media undercounts police (and organization) estimates (13 percent). Finally, when organizations expected more participants than the police counted, media tend to report higher demonstration turnouts than police counts in no less than 61 percent of the events. This adds strong evidence against the perspective that media blindly follow police counts and undercount demonstrations. Again, the marginalization cell is the least populated cell of the row (9.1 percent, n = 3). In sum, Table 4 presents little evidence of journalists systematically underreporting organizer expectations or of journalists blindly following police counts. The fact that the row percentages across the columns follow a logical pattern depending on the relationship between organization expectation and police count (decreasing in the “Media < Police” column, increasing in the “Media > Police” column) suggests that journalists react quite rationally and reasonably to differential guesstimates rather than systematically marginalizing protest. A final argument against a logic of marginalization is that for only 8 out of the 106 demonstrations in Table 4 (7.5 percent) media report a number lower than both police and organizer guesstimate (result not in table).

We now turn to explaining tilt in turnout coverage. When is factual turnout presentation more likely to play down police (Model 1) or organizer (Model 2) guesstimates? And, when are journalists more likely to present the average of both (Model 3)? Table 5 presents the analyses, with significant odds ratio’s printed in bold (p < 0.05).

Table 5.

Logistic Regressions Predicting Undercounting and Presenting the Average.

	Model 1		Model 2		Model 3
	Undercounting Police Count		Undercounting Org. Count		Presenting the Average
	Odds	Sig.	Odds	Sig.	Odds	Sig.
Marginalization
Union	0.392	.019	0.644	.339	0.827	.658
Human rights issue	0.811	.661	2.294	.194	0.975	.964
Asylum issue	1.106	.827	1.906	.285	0.259	.064
Peace issue	0.963	.954	1.732	.545	1.149	.843
Environmental issue	0.344	.223	1.654	.608	0.391	.418
Normalization
Large turnout	0.355	.003	0.752	.479	2.267	.022
Guesstimate Gapa	1.082	.100	0.575	.000	0.659	.000
Controls
Lead item	0.689	.499	0.452	.238	1.837	.282
Public broadcaster	1.738	.091	1.005	.990	0.810	.547
Constant	0.415	.134	0.768	.710	1.070	.924
Nagelkerke R2	.152		.531		.165
N	217		217		217

^aIn case of Model 3 (Average), we use the Abs. Guesstimate Gap variable.

Note. Bold signifies values below .05.

H3 expected downplaying of guesstimates to be a function of issues and organizations. H3 gets little support from the data. Across the three models, only trade union staged protest is significantly different from other protests: Factual turnout coverage is less likely to underreport the police estimate of trade union protest. Again, more powerful explanations of turnout coverage lie in the size-related variables. Turnout coverage of large demonstrations is less likely to undercount the police estimate (Model 1), whereas demonstration size is not related to how media cover the organizer’s expectation (Model 2), rejecting H4. Interestingly, if the guesstimate gap increases—police counts more participants than the organizers—media are less likely to present a turnout that is lower than the organizer’s estimate (Model 2). Both results suggest a certain stickiness of journalists to the police count, adding weight to H5.

To get further to the bottom of tilt in factual turnout coverage, we ran several multinomial logistic regressions (not in table) using media downplaying, reflecting, and up playing as the categories in a single dependent variable. These results further confirm the stickiness of police counts. Turnout presentation of large demonstrations is less likely to undercount police estimates (while having no effect on organization estimates). And, if the police counts more participants than the organizer expected, turnout in the media is more likely to be larger (and less likely to be lower) than the organizer’s expectation. The opposite holds as well: If organizers expected more participants than counted by the police, the turnout as mentioned in the media is more likely to be lower that the organizer’s expectation. In addition, these analyses reveal that turnout of antiwar demonstrations was more likely to be played up in the media arena (police estimate as baseline), a finding that reconfirms the exceptional situation of media coverage of Iraq war protests.

Finally, when it comes to presenting the average of police and organizer guesstimates (n = 24), the evidence shows that journalists do so more frequently in case of large demonstrations (Model 3). However, odds of journalists taking a middle ground position decreases when the difference between organizer and police count grows. This negative effect of the guesstimate gap variable is even stronger for large demonstrations (interaction term “Large demonstration × Abs. Guesstimate gap”; odds = 0.583, Sig. = .015). Journalists thus are more likely to present the average if guesstimates are relatively close to each other, and hence, the distorting effect of reporting an average is relatively small.

All said, when it comes to our second block of hypotheses, we need to reject H3 (tilt in coverage is not a matter of issue or organizations) and H4 (large demonstrations are not more likely to be played down), but have some evidence that confirms H5: Police estimates strongly influence the figure journalists report (Table 5), but journalists do not blindly follow police estimates (Table 4).

Discussion and Conclusion

Size matters in politics. Whether it is votes in elections, attendance at inaugurational speeches, or protesters marching in the streets, power is in numbers. In this paper, we systematically scrutinized media coverage of protest size, relying on a unique dataset combining multiple turnout perspectives. We created a typology, investigated drivers of turnout coverage, and compared media turnout with police and organizer guesstimates. Together, these analyses aimed at a larger, looming question relating to issues of bias, power, and control in the media arena. Is media coverage of protest size a matter of marginalization or normalization? Let us first systematically revisit the findings of this paper. Our results show the following:

• Three news items in four present information on demonstration turnout (Table 2).

• Ninety percent of the news items reporting turnout present a single turnout figure (factual turnout coverage); 10 percent of the news items present multiple counts or interpret turnout (contentious turnout coverage) (Table 2).

• Large demonstrations and demonstrations with a larger gap between organizer and police estimates are more likely to receive contentious turnout coverage. Issues and organizations are far less potent predictors of how turnout is covered (Table 3).

• In case of large demonstrations, media are unlikely to present a factual turnout figure that is smaller than the police count (14 percent). They reflect police counts (44 percent) or present a higher count (42 percent). Media do not blindly follow police counts (Table 4).

• Police estimates and, more specifically, their relationship to organizer expectations (does the police count more attendants compared with the organizer’s expectation or less), strongly influence how media report turnout and mold organizer guesstimates (Table 5).

• Finally, journalists present a middle ground position more frequently in case of large demonstrations, but less so if the gap between organizer and police guesstimates increases.

What are the implications of these results? Is turnout coverage a matter of marginalization or normalization? In our view, the evidence overwhelmingly supports the latter perspective. Turnout coverage in our case appears to be by and large a function of event size characteristics. It is turnout cues presented by the organizers and the police that are most potent in shaping turnout coverage, not the type of organization or the issue tackled by the demonstrators. If demonstrations are large, or if guesstimates diverge, journalists pay more attention to turnout by giving multiple counts or by putting turnout in perspective. In our view, this attests of professional journalists who react to event cues, not of status quo protectors or partisan journalist with an agenda.

Several critical observations can be made against our interpretation. First, organizer expectations and police estimates are rarely similar, yet journalists most often present a single count (90 percent). The rarity of contentious turnout coverage is surprising. Do journalists do so to obscure potential debate? We think not. Factual turnout coverage is more likely for small demonstrations, when guesstimates are closer to each other and for items that do not make headline news. These are the less newsworthy and impactful events for which turnout is far less key. It is very likely that mundane aspects, like time constraints of the newscast, explain the infrequency of contentious turnout coverage. Additional analyses show that factual turnout items are significantly shorter than contentious ones (M_cont = 246 s; M_fact = 97 s), t(414) = −10.533, p = .000), adding some empirical flesh to this interpretation.

Second, police counts seem to set the boundaries of how journalists mold organizer expectations in case of factual turnout coverage. This finding resonates with rationales on media’s focus on official sources, and how this focus reinforces the status quo. However, our results also show that media do not blindly follow police counts; and, if guesstimates diverge, contentious coverage becomes more likely. Also, the superiority of police counts in explaining factual turnout might simply have to do with the superiority of the measure. An obvious downside of this paper is that we compared organizational expectation before the demonstration with police guesstimates after the demonstration. Obtaining systematic data on organizational assessment of crowd size after the demonstration would strongly improve our design. Finally, the fact that we consider turnout coverage to be normalized does not suggest that making the news is easy for social movements. Eliciting media attention remains very much an uphill struggle for social movements; most protests do not get any media attention at all. Whereas the odds of attracting the media spotlight are quite bleak indeed, our results show that if protest gets covered, the representation of its size attests more to a logic of normalization than marginalization.

Our results bring up important questions about generalizability. Our findings hold for television news in a democratic-corporatist media system with a strong public television station. Like other West European countries, the Belgian media environment is characterized by organized and professionalized journalists and the state as a facilitator of the information flow. We believe our results to apply to similar news ecologies. Other media systems, however, with more partisan press and narrowcasting television stations—for instance with Fox News and MSNBC in the United States—are very likely to report differently on similar demonstrations (Weaver and Scacco 2013). In such cases, it might be not so much event characteristics that determine how turnout is reported. Rather, aspects of issues and organizations in interaction with the political color of a medium might drive turnout presentation in these instances. Besides a marginalizing effect (undercounting), also a rally around the flag-effect then might be at play, amplifying the significance of ideologically close movements and further polarizing politics. The medium, finally, might impact turnout coverage as well. Television news has a small news hole and can present many cues about demonstration turnout visually. It might well be that turnout coverage in newspapers, with less possibilities to share visuals but more space to elaborate on protest size in text, presents turnout more contentiously.

In sum, the typology we presented here begs for replication and application. Replication, preferably in a comparative design (see, for instance, Shahin et al. 2016), would allow addressing questions related to the robustness and contingencies of our case results. Application, second, could lie in testing the effects of different types of turnout coverage on audiences. Especially contentious coverage (interpretation and contestation) and its respective tilts ([which source presents a] lower or higher turnout) could affect how news consumers perceive the potency of protest. Our study also made us wonder what type of turnout coverage—under what conditions—would be considered in line with standards of “good journalism.” In-depth interviews with journalists on how they cover demonstration turnout could shed light on the process of how journalists construct turnout in the news, an issue unaddressed in this paper.

In closing, in times of mass media echo-chambers and social media filter bubbles (Jamieson and Cappella 2008; Pariser 2011), professional journalism and its standards are under pressure. Citizens meet mass media with distrust; politics is labeled “fact-free” or even “post-truth.” The results in this paper show that journalists in a democratic-corporatist media system working for broadcasting television stations cover groups that seek social change in reaction to event cues, regardless of the issue that is being tackled or the organization that is involved. Whether this finding holds in other contexts and what elements exactly condition such coverage, should be of interest to scholars studying media–politics relationships.

Author Biographies

Ruud Wouters is an assistant professor in Political Communication & Journalism at the University of Amsterdam. His research interests are media coverage of social movements, the impact of protest, and protest participation.

Kirsten Van Camp holds a PhD from the University of Antwerp and is currently working as a researcher at the Political Science Department there. Her main research interests include media coverage of political elites, issue ownership and the journalistic news production process.