Demonstrating Power: How Protest Persuades Political Representatives

ASRXXX10.1177/0003122417690325American Sociological ReviewWouters and Walgrave 6903252017

American Sociological Review 2017, Vol. 82(2) 361–383 Demonstrating Power: How © American Sociological Association 2017 https://doi.org/10.1177/0003122417690325DOI: 10.1177/0003122417690325 Protest Persuades Political journals.sagepub.com/home/asr Representatives

Ruud Woutersa and Stefaan Walgravea

Abstract How do public opinion signals affect political representatives’ opinion formation? To date, we have only limited knowledge about this essential representative process. In this article, we theorize and examine the signaling strength of one type of societal signal: protest. We do so by means of an innovative experiment conducted among Belgian national and regional politicians. Elected officials were exposed to manipulated television news items covering a protest demonstration. Following Tilly’s previously untested WUNC claim, four features of the event were manipulated: the demonstrators’ worthiness, unity, numerical strength, and commitment. We argue that these protest features present elected officials with useful cues about what (a segment of) the public wants. We find that these cues affect elected officials’ beliefs. The salience they attach to the protest issue, the position they take, and their intended actions all change as a consequence of exposure. The size of a protest event (numbers) and whether the protesters agree among themselves (unity) are the most persuasive protest factors. The effects of the protest signals come on top of strong receiver effects. We find no evidence that elected officials’ predispositions moderate the effects of the protest features.

Keywords protest, representation, opinion formation, elected officials, experiment

In a democracy, elected representatives are To date, we have only limited knowledge supposed to represent society. In the words of about this essential representative process of Dahl (1973:1), “a key characteristic of a individual opinion formation by elected offi- democracy is the continuing responsiveness cials (for a similar argument, see Belchior of the government to the preferences of its 2014; Broockman and Skovron 2013; Miler citizens.” Because elected officials aim to 2007). To be sure, we do know that elected remain in office and fear electoral sanctions, officials resort to different sources, like they have an interest in keeping close track of what the public wants and are expected to adjust their beliefs and actions accordingly aUniversity of Antwerp (Downs 1957; Miller and Stokes 1963). But how do elected representatives form percep- Corresponding Author: tions about what the public, or a relevant seg- Ruud Wouters, Research Group Media, Movements & Politics (M²P), University ment of the public, wants? And, most impor- of Antwerp, Lange Nieuwstraat 52, 2000 tantly, to what extent are their own opinions Antwerpen, Belgium affected by these perceptions? E-mail: [email protected] 362 American Sociological Review 82(2) opinion polling (Geer 1996), mass media opinion of a segment of the public. The sig- (Herbst 1998), contact with constituents nals we confront them with are different (Fenno 1978), and different forms of advo- media portrayals of a protest event. We test to cacy (Burstein 2014) to gauge what the public what extent different versions of the signal— wants. And a large and growing body of in this case, carefully manipulated features of research scrutinizes the actual correspond- the protest action—differently affect the ence between public opinion signals and pol- extent to which representatives align their icy behavior; that is, what elected officials do own opinion with that of the protesters. (Burstein 2014; Erikson, MacKuen, and Access to a large sample of national elected Stimson 2002; Page and Shapiro 1992; officials and the assets of experimentation Soroka and Wlezien 2010; Wlezien 2004). give us unique leverage to address questions These studies find considerable evidence for of how incoming societal signals affect repre- a link between public opinion and policy. Yet sentatives’ opinion formation. this general pattern hides a great deal of vari- We consider protest as one particular pub- ation in responsiveness as well. Policies tend lic opinion signal that may influence elected to better reflect the preferences of the most representatives. Politicians, of course, are affluent, for instance (Bartels 2008; Gilens exposed to many other societal signals. 2005). Citizens with more money and politi- Focusing on protest as an expression of a part cal skill can better organize, have better of public opinion has distinct substantive, access to policymakers, and manage to com- theoretical, and methodological advantages. municate their interests better and with more Substantively, staging protest is the weapon pressure (Schlozman, Verba, and Brady of the weak. Protest groups typically are 2012). Similarly, public policy is biased resource poor. They lack direct access to the toward the status quo, protecting the powers policymaking process and often challenge that be against the efforts of those who chal- (the lack of ) extant policies. These elements lenge extant policy (Gilens 2005). Finally, make responsiveness unlikely. By staging democratic linkage is strongest for issues the protest and going public, protesters seek general public cares most about (Burstein social support, hoping to set in motion a pro- 2014; Wlezien 2004). Politicians’ leeway to cess that will make it hard for elected officials cater to special interests is considerably larger to continue ignoring them. As such, studying for (the many) issues that stay under the pub- the persuasiveness of protest signals is of lic radar. particular substantive interest, as it can In summary, although we know that politi- increase our understanding of the conditions cians resort to different sources to learn about under which a democracy can be more inclu- public opinion, and although we have sub- sive (or at least pluralistic). Focusing on pro- stantial evidence about democratic congru- test as societal signals also has methodological ence and its contingencies, little is currently and theoretical advantages. Protest signals are known about how individual elected officials clearly delineated in terms of place, time, and process incoming public opinion signals and issue and can thus be studied more easily than how these signals affect their own opinions. broader, more ambivalent signals. Addition- To further our knowledge about how ally, we can build on a sizable body of exist- elected representatives form opinions—opin- ing theory about when protest matters (for a ions that may later affect the actual policy review, see Amenta et al. 2010; Andrews initiatives they take or endorse—this study 2004; Biggs and Andrews 2015). draws on an innovative experimental design More concretely, we theorize that who the and an exceptional set of respondents. We protesters are, how they behave, and what examine the process of opinion formation by they think determines the strength of the pro- elected representatives by presenting them test signal and affects the calculations elected with vignettes containing a story signaling the officials make on whether and how to Wouters and Walgrave 363 incorporate the protest signal in their own The Importance of thinking (Gillion 2013; Lohmann 1993). In Elected Representatives’ particular, we test Tilly’s (2004) widely cited Opinions but untested claim that protest influences power-holders when protesters display wor- Ultimately, the political outcome of any soci- thiness, unity, numbers, and commitment etal signal to power-holders depends on the (WUNC). We conducted a survey-embedded perceptions, calculations, and subsequent vignette experiment with a large sample of actions of elected representatives (Jones, elected politicians in Belgium. These repre- Boushey, and Workman 2006). In the chain sentatives were exposed to manipulated tele- from signal to representation, the attitudes and vision news items of a protest action regarding beliefs of individual politicians matter for the asylum issue. Asylum is a heavily politi- their political decision-making and for poli- cized issue worldwide. In Europe in particu- cies in general (Carnes 2013; Mansbridge lar, it has been high on both the public and 1999; Mondak 1995). Also with regard to the political agenda since the onset of the Syrian impact of protest and social movements, the refugee crisis. In our study, fielded well idea that individual politicians’ opinions mat- before the dramatic increase in refugees ter has gained ground (Burstein 1999; Luders requesting asylum in Europe, the demonstra- 2006; Skrentny 2006). In fact, the social tors criticized the Belgian asylum procedure, movement literature holds that the presence of demanding a more humane process in due “elite allies” is an important, maybe even the time based on clear regularization criteria. In single most important, factor of movement the news item stimuli, the WUNCness of the success (Amenta, Caren, and Olasky 2005; asylum protest event varied systematically. Amenta, Carruthers, and Zylan 1992; Cress After exposure, representatives were sur- and Snow 2000; Soule and Olzak 2004). veyed about their own opinion regarding the We argue that at least three aspects of salience of the protest issue, their position on elected representatives’ opinions are relevant the protest issue, and the actions they intended and may affect subsequent actions: (1) the to undertake regarding the protest issue. salience representatives attribute to an issue; Results show that politicians do react dif- (2) the position they hold on the issue; and (3) ferently to different portrayals of protest. their intention to undertake action regarding Elected representatives’ own opinions regard- the issue. These three dimensions of elected ing the protest issue and the extent to which officials’ opinions can be considered condi- they adopted the claims made by the protest- tions for their actual action to be in line with ers varied across manipulations of the protest the expressed public opinion. signals. Displays of unity and the number of First, for a politician to act, she needs to protestors involved were the most influential care about the issue and find the problem protest features shaping representatives’ own important or salient. We know that attention issue salience, position, and intended action is a scarce resource in politics (Hilgartner and beliefs. Worthiness and commitment induced Bosk 1988). Moreover, a considerable litera- effects on position and salience, respectively. ture in political science shows that decision These signal effects came on top of elected makers’ attention is an absolute precondition officials’ strong preexisting beliefs and opin- for political change (Baumgartner and Jones ions regarding the issue at stake. We conclude 1993; Jones and Baumgartner 2005). Politi- that the type of societal signal matters for cians are bombarded with information and representatives’ opinion formation. Particular cannot attend to all incoming signals. A signal features of protest can push elected repre- about public opinion thus needs to stand out sentatives’ opinions to become more (or less) to get noticed. If elected officials’ perceptions in line with the public opinion signals they of the importance or urgency of an issue is are exposed to. positively affected by a societal signal, the 364 American Sociological Review 82(2) odds that they will actually intervene increase. In summary, elected politicians’ opinions This general logic also applies to protest sig- are important because they may foreshadow nals. Protesters’ primary aim, often, is to political action and actual policy. This makes simply draw political attention to a problem the process of representatives’ opinion forma- (King, Bentele, and Soule 2007; Walgrave tion central to representation. and Vliegenthart 2012). Second, for a politician’s potential action to be in line with what the public wants, she How Elected needs to more or less substantively agree with Representatives the direction of public opinion expressed in Form Opinions the signal. In other words, the politician should hold a position that is in line with the defini- Political representatives are in the business of tion of the problem put forward and see some having opinions about societal matters. Politi- merit in the solution encapsulated in the sig- cians want to change society according to nal. A classic literature in political science their values and principles; they aim to imple- describes the positioning of elected officials ment their plans for what they consider to be (going back to Downs 1957) and how political a better society (Strøm 1997). So, politicians’ representatives’ positions relate to those of opinions about issues are likely to be quite their constituents (Miller and Stokes 1963). fixed and directly connected to their ideolo- This literature holds that constituents and rep- gies (Uslaner 1999). Still, elected officials’ resentatives having similar positions increases opinions are not necessarily entirely stable. the chance of adequate representation (Erik- Their opinions can evolve in reaction to son et al. 2002; Soroka and Wlezien 2010). external stimuli, particularly signals about Applied to protest signals, the more repre- citizens’ preferences, as neglecting these sig- sentatives agree or sympathize with a protest’s nals may endanger re-election. demands, the higher the chance politicians According to democratic theory, political will act in line with what the protesters want. representatives’ main motive is to secure re- Third, finding an issue important and agree- election (Downs 1957). Even when repre- ing with the position of expressed public opin- sentatives aim to develop policy, getting ion is not enough—politicians must also intend re-elected is the crucial precondition (May- to act on these beliefs. Politicians have a whole hew 1974). As a consequence, politicians repertoire of different types of political action need to maximize votes, and to that end they at their disposal. They can take a public stance, need information about citizens’ preferences try to convince their fellow partisans to act, (Erikson et al. 2002). To stay in tune with the introduce a bill, ask a question in parliament, electorate and minimize the risk of being and so forth. All these actions are planned and blamed at election time (McGraw 1990; carefully deliberated. Psychologists contend Weaver 1986), politicians pay attention to that people rarely act unless they intend to do many informational cues, ranging from pub- so (Fishbein and Ajzen 2011). Thus, intended lic opinion polls and research reports to media action is the third precondition for an individ- coverage about the issues of the day (Herbst ual politician to act in line with public opin- 1998). All these sources allow elected repre- ion—in our case: acting according to a protest sentatives to track public preferences. Note signal. Of course, there is a gap between what that representatives are not expected to weigh people say or intend to do and what they actu- all segments of the public equally. They are ally do (and this is certainly true for politi- expected to devote special attention to citi- cians!). But examining intended behavior is an zens living (and voting) in their electoral often-used strategy when directly observing district, to voters supporting their party, to (or eliciting) behavior is impossible or ethi- parts of the population they explicitly care cally reprehensible. about because of ideological reasons, or to Wouters and Walgrave 365 segments of the public that have the resources media connects specialist publics with the to put them under pressure (e.g., by sponsor- general public, so information and debates ing their campaign, or not). presented in this forum can reach inattentive We argue that representatives’ exposure to bystander publics, possibly turning these pas- societal signals affects not only their percep- sive spectators into active stakeholders (Fer- tion of what the public wants but also their ree et al. 2002). According to Key’s (1961) own opinion. Because political survival classic account of public opinion, it is the through public support in elections is a repre- uncertainty about this activation of latent sentative’s main motive, it makes sense for public opinion that causes most anxiety among them to let their own opinion be affected by political representatives. Arnold (1990), simi- their perception of what the public, or a rele- larly, argues that legislators constantly antici- vant segment of the public, wants. Whether pate future public preferences and involvement representatives are right in assuming that the pub- when making decisions, estimating citizens’ lic will punish them for acting non-responsively potential preferences and the electoral “bite” and will reward them for acting responsively of an issue, out of fear of being held account- does not really matter (Lang and Lang 2008). able at election time. As long as they think their behavior is visible In this complex information environment, and the public will react to it, they are likely to protest signals form a rare species of relatively update their own preferences to bring them in succinct, clear, and manageable information. A line with what they think the people think. In protest signal yields cues about (1) how many other words, we expect political representa- people and (2) which people care about an tives, to some extent, to align their own opin- issue, (3) to what extent they care, and what ion with the public opinion signal coming in their exact (4) position on the issue is. Of from society. course, protest signals are unclear to some The protest we examine in this study is just degree, because it may be in protesters’ inter- another public opinion signal politicians get ests to mislead political decision makers. As from society. Protest actions provide repre- Burstein (1999) notes, groups who believe sentatives with specific bits and bytes of their views to be supported by a popular major- information that improve their understanding ity want politicians’ perceptions to be accurate. of what the public, or a specific segment of In contrast, groups whose preferences are in the public that they particularly care about, the minority want their positions to be per- wants (Gillion 2013; Lohmann 1993). The ceived as more popular than they actually are. attractiveness of the information encapsulated In any case, bringing an issue out in the in protest can best be understood when look- open by means of protest can influence ing at representatives’ information environ- elected officials directly, by changing their ment. In this environment, information is perception of what the public wants (be it abundant and acquiring accurate information accurate or not), or indirectly, by setting in is costly and time intensive (Miler 2009; motion and activating public opinion itself, Simon 1962). In addition, the accuracy of the which in turn can put pressure on representa- public opinion information representatives tives, either in line with what the protestors receive is often problematic. Public positions want, or against their claims, by encouraging and priorities may be volatile (Zaller 1992). counter-movement mobilization (Burstein Representatives’ knowledge about what the 1999; McAdam and Su 2002; Zaller 1992, public wants is thus always imperfect at best 1997). All these pathways of potential impact (Miller and Stokes 1963). The uncertainty of come together in the mind of the ever-antici- the information environment is further ampli- pating politician, affecting her calculations fied by the role of mass media, which oper- when she decides whether and, if so, how to ates according to its own logic (Blumler and incorporate the incoming information and Kavanagh 1999; Strömbäck 2008). Mass align her own opinion with it. 366 American Sociological Review 82(2)

Our account of protest events as informa- issue and will not give up easily (commit- tive signals is not entirely new (see Gillion ment). These four elements form an implicit 2012; Lohmann 1993; McAdam and Su 2002), scorecard against which observers assess but our strong emphasis on the receiver side of demonstrators’ strength. They nicely map the signal is. Previous work in social move- onto the public opinion cues politicians are ment studies has considered protest actions as looking for. WUNC matters because it makes communicative acts (Etzioni 1970), examined a powerful assertion of popular sovereignty the importance of protest being projected and (Tilly 2004:13): the protesters are serious perceived (Lipsky 1968), discussed the rele- claimants, who legitimately speak on behalf vance of gaining (supportive) media attention of society, and who can use their strength to (Koopmans 2004), and investigated the related “enter, realign, or disrupt the existing polity” mechanism of scope enlargement (Gamson (Tilly 1999:262). From the perspective of 2004). All these accounts are implicitly based political representatives, high WUNC pro- on the idea that protest forms a public opinion tests broadcast the existence of a potential signal that may affect representatives’ percep- voting block that might influence their tion of public opinion, and subsequently, their chances of re-election (Tilly 2004). own opinion. Note that Tilly talks about WUNC “displays” (Tilly 2004:4) and “broadcasting” WUNC (Tilly 1995:373). WUNC is a perfor- What matters in A mance, something protesters deliberately protest signal? seek to portray. It is the result of “mystifica- What exactly in a protest triggers elected offi- tion” processes and more a matter of appear- cials to update their beliefs about the protest ance than of reality (Tilly 1994:16). Given issue? Where does the strength of a protest that WUNC is a subjective construction by signal stem from? We argued that representa- protesters seeking to persuade observers, the tives are looking for cues about (1) how many mass media arena emerges as the main stage people and (2) which people care about an on which contemporary movements seek to issue, (3) to what extent they care, and (4) enact WUNC (Koopmans 2004). Tilly (2008: what their exact position on the issue is. In 74) considers protest demonstrations in par- this section, we argue that specific features of ticular as vehicles to show off WUNC, a protest event may provide representatives because they are like “miniature social move- with exactly these cues (for a similar arguments” and “nicely encapsulate the distinc- ment, see Gillion 2013; McAdam 1996). We tive features of social movement displays” expect these protest features lead representa- (Tilly 1999:260). tives to update their opinions. Tilly never explicitly operationalized the The most extensive claim of how protest four WUNC elements beyond exemplary includes public opinion signals that may descriptions (see, e.g., Tilly 2004:4, 2006:291) affect observers was formulated by Tilly in a nor connected the individual WUNC ele- series of publications (1994, 1995, 1999, ments to related strands of research. We do 2003, 2004, 2006, 2008, 2010). According to this here and present our interpretation of Tilly, political representatives are responsive Tilly’s WUNC features. to protests when protesters display worthiness, unity, numbers, and commitment (WUNC). Worthiness. Tilly’s first feature of protest Although Tilly’s WUNC idea rarely took matches what Schneider and Ingram (2005) center stage in any of his work, we believe the call the “deservingness” of the beneficiaries framework has merit and use it as our point of of public policy. By behaving in a worthy departure. WUNC signals that protestors are fashion, protesters signal to politicians that credible (worthy), are a large group (num- they are good citizens and that the protest’s bers), agree among themselves about what claim is supported by a segment of the public they want (unity), and strongly care about the that deserves to get what it wants because it Wouters and Walgrave 367 behaves in an appropriate manner. In other uniform gestures (applauding, chanting, words, worthiness provides representatives walking together) and symbols (badges, with information about what kind of segment colors, flags) (Tilly 2004). Unified protesters of the public the protest signal comes from. should be more effective in persuading Specifically, within the field of social elected officials because of two reasons. First, movement studies, the worthiness feature can unified protestors produce a clear signal, be related to the ongoing disruption-modera- which is easier to interpret and follow up. tion debate (Giugni 1999). According to Tilly, Protests with multiple claims or whose pro- movements sacrifice the advantages of vio- testors disagree, send ambiguous, messy sig- lent action and choose to behave non-vio- nals; it is unclear what exact position they are lently to gain recognition as respectable defending and they leave politicians puzzled. players who should be listened to. Tilly hints Second, message coherence may be an indi- at the non-violent nature of worthy protest by cator of organizational strength of the rele- describing it as “eloquent” (Tilly 1994:13), vant segment of the public. Protestor unity “disciplined” (Tilly 2008:144), and not “dis- can give politicians the impression of a well- reputable” (Tilly 2006:291). Tilly’s stance on organized actor who can mobilize against non-violence is provocative because of a lack them, or whom they can rely on and work of empirical consensus. Early work on the with to strike a bargain. impact of social movements by Gamson Within the field of social movement stud- (1990) and especially Piven and Cloward ies, the relevance of unity can be traced back to (1993) emphasized the advantages of disrup- theories of frame alignment (Benford and tive strategies. The mechanism behind Snow 2000). Yet, most framing studies focus responsiveness to unworthy protestor behav- on mobilization, and only a few have tackled ior is that officials “trade concessions for the effect of message coherence on political tranquility” (De Nardo 1985:35; McAdam outcomes of protest. Gamson (1990), in his 1983). An overview of protest impact studies seminal study, found single-issue groups were by McAdam and Su (2002), however, con- more successful than multi-issue groups. Cress cludes that the relationship between disrup- and Snow (2000), similarly, demonstrated the tion and policy outcomes is mixed or effectiveness of more articulate and coherent inconclusive at best. With the normalization frames in their study on the mobilization of the of protest in Western democracies (Dalton homeless. More recently, Fassiotto and Soule 2008; Meyer and Tarrow 1998), people have (2015), relying on a category-spanning frame- come to see peaceful protest as a legitimate work (Hsu, Hannan, and Koçak 2009), showed way of making demands (Norris, Walgrave, how women’s protests with focused messages and Van Aelst 2005), whereas the opposite were more likely to raise congressional atten- has happened with regard to disruptive con- tion compared to protests sending more muddy tention (della Porta and Diani 1999; Tarrow signals. In summary, we expect clear, focused 1994). Violent behavior is therefore likely to protest messages to increase representatives’ lead to marginalization and even criminaliza- responsiveness to a protest signal. tion of protestors, and it can alienate elected officials and potential sympathizers. In sum- Numbers. Demonstration turnout as a mary, we expect violent demonstrators to source of protest strength is most directly negatively influence elected officials’ opinion related to theories of democratic representa- of a protest issue, because the violence can tion. De Nardo (1985) and della Porta and influence officials’ perceptions of protesters’ Diani (1999) both hold that “power is in num- worthiness. bers.” Numerical strength aligns with the majoritarian logic of representative democ- Unity. Unity is a matter of message coher- racy. Protest size gives elected politicians ence and displaying this coherence by cues about the broader support demonstrators 368 American Sociological Review 82(2) enjoy in society, and hence about the size of expect elected representatives to update their the segment of the public involved (Burstein opinions accordingly. and Linton 2002; Lohmann 1993). The larger In summary, in line with Tilly, we expect the protest the higher the chance that the all four WUNC features to yield positive majority of the public shares the protesters’ effects on elected officials’ opinion forma- views, causing representatives to cater to tion. Note, however, that our literature review demands voiced by large demonstrations. also adds some relief to this general expecta- Again, although the relationship between tion. Whereas the impact of numbers can be demonstration size and demonstration out- considered obvious, given its match with the comes is plausible, few protest studies directly logic of representative democracy, extant lit- tackle this puzzle. McAdam and Su (2002) erature is not on the same page about the show how larger peace demonstrations effect of a feature like worthiness. increased the pace of congressional action during the Vietnam War. Walgrave and Vliegenthart (2012) find that especially large Data and Methods demonstrations move legislation across a We fielded a survey-embedded TV-vignette wide range of issues. In summary, we expect experiment among Belgian elected officials elected officials’ opinions to be susceptible to from March 2015 to May 2015, well before the cues about protest size, with large demonstra- outburst of the Syrian refugee crisis.1 Vignette tions positively affecting elected officials’ designs confront respondents with fictional opinion formation. descriptions of a situation. Several characteris- tics of this situation are systematically manipu- Commitment. According to Tilly (1994), lated (independent variables) to tease out the demonstrators show commitment by persist- extent to which these aspects play a role in ing in costly activity and showing readiness forming judgments (dependent variable). For to endure. Committed protestors convince an overview of the use of vignette studies in elected representatives that the activity is not sociology, see Wallander (2009). simply a fad but that dissatisfaction is deeply rooted. Because committed people pursue Procedure their goals at the expense of other potential activities and interests (Hunt and Benford The respondents (N = 269) for this study are 2004), committed protestors will likely per- all Dutch- or French-speaking elected national sist and vote accordingly. The prospect of (42 percent) or regional (58 percent) politi- continuous contention, which might raise the cians in Belgium. The target population con- salience and electoral bite of an issue, may sisted of all 370 Members of the Belgian make politicians responsive to such “issue national parliament, the Flemish regional publics” (Popkin 1991). parliament, and the Brussels-Wallonia Agenda-setting studies of protest impact regional parliament (251 MPs participated). have measured commitment by means of This sampling frame was complemented by demonstration frequency. Although some adding the party leaders of all parties sitting studies find no or limited evidence for the in one of these parliaments (six party leaders effect of protest frequency (Giugni 2007; participated) and the ministers in the respec- Olzak and Soule 2009), other studies do find tive national and regional governments (12 effects (Johnson, Agnone, and McCarthy ministers participated). In total, 65 percent of 2010; King et al. 2007; Walgrave and all contacted politicians participated in the Vliegenthart 2012). Because displays of com- study—a high response rate for a target group mitment by protesters encapsulate a strong known to be notoriously pressed for time (for cue about how much a part of the public a study among MPs in 15 countries, with a really cares about the underlying issue, we response rate ranging between 12.9 and 43.3 Wouters and Walgrave 369 percent, see Deschouwer and Depauw 2014). attentively. After each item, they were given a This high degree of cooperation might be due, number of questions tapping their beliefs in part, to Belgian exceptionalism, the conse- about the protest issue and the protestors. quence of a political culture in a small Clips were grouped in sets of two, with each democracy where elected officials cooperate clip being the complete opposite of the other, in academic research. Our way of approach- making for a balanced design (Atzmüller and ing elected officials, in tandem with the sub- Steiner 2010). Sets were randomly distributed ject of the broader study, also boosted the across respondents, and the clip order within response rate. We contacted elected officials a set was randomized. Dutch and French clips with an official letter that explained the pur- were identical in terms of footage and voiceover, pose of the study (how politicians tame the only voiceover language differed. The results information tide and how this influences their of a manipulation check showed the news job as representatives), mentioned the fund- items were perceived as realistic and the ing agency, stressed why cooperation mat- manipulations came across as intended. Full tered, promised anonymity, and gave the details on the construction of the TV clip option to receive a descriptive study report. stimuli, exact question wording, randomiza- The letter closed by stating that in the follow- tion, the manipulation check, and basic ing week(s), the elected official would be descriptives can be found in the Appendix contacted by telephone by the principal inves- and the online supplement. tigator of the project to answer all possible questions and to schedule a meeting of Independent Variables: Protest approximately one hour, consisting of a sur- Features in TV News Vignettes vey and an interview. Elected officials were re-contacted until an appointment was made We investigate the influence of four protest or until they explicitly refused cooperation. features, the key independent variables of the Political representatives were surveyed by a study: worthiness, unity, numbers, and com- team of trained and experienced researchers mitment. We manipulated a fifth protest fea- who visited them in their Brussels offices. ture (diversity) when creating the vignettes, Respondents completed the survey on a lap- but given that diversity was not part of Tilly’s top provided by the interviewer. original theory, the non-significant effect of The protest vignettes were shown at the the diversity manipulation, and space con- end of the longer (35 minutes) survey that straints, we do not report the diversity results dealt more generally with elected officials’ here. Because of the randomized exposure information-processing behavior. The survey and balanced design, this has no effect on the started with questions tapping the sources of results. information the politicians attune to, ques- All manipulated protest features could tioned them about issues they considered have two values (high or low), so we created important, and asked them to engage in a 32 (25) different clips, each containing a process-tracing exercise about their most unique combination of all five protest fea- important political initiative. We have no rea- tures (four of which we report here). The son to believe that the broader set-up of the protest covered in the news item dealt with survey had any confounding effect on our the Belgian asylum procedure. Asylum is an results. Politicians were shown two fictional important and heavily politicized issue world- television news reports about a demonstration wide and particularly in Europe and Belgium. on the same topic. The TV news report lasted The realism of the topic and the setting in about one minute, was voiceover only, which the experiment was conducted ensured showed the logo of the public broadcaster, that elected officials engaged in an authentic and dealt with the issue of asylum-seekers. way with the stimuli (see the following sec- Politicians were asked to watch both items tions). The external validity of the experiment 370 American Sociological Review 82(2) is further increased by the fact that asylum- single message, which was crystal clear: the seekers frequently demonstrate in Brussels. asylum policy must become clearer and One of the study authors collected police more humane.] [Worthiness: Towards the archive data on demonstration activity in end of the demonstration, the atmosphere Brussels showing that, between 2001 and turned grim. Several shop-windows were 2010, at least 446 protest events were staged broken and demonstrators started pushing on the issue of asylum and migration. The and pulling. A group of troublemakers was same protest event analysis found that asylum- arrested.//The demonstration went off with- seeker demonstrations in Belgium vary out incidents and in a serene atmosphere, strongly on the WUNC elements: there are with demonstrators behaving calmly and small and large events, violent and non-violent peacefully.] [Commitment: Whether further ones, and these demonstrations regularly actions will follow is unclear. At the moment make it to the television news (Wouters no new demonstrations are planned.//After 2013). This underscores the mundane realism the demonstration in Antwerp ten days ago, of representatives being confronted with TV this action is already the second in a row. A news about different asylum protests. The follow-up demonstration is planned for next demonstrators’ precise claim was featured in week.] the introduction of the news items and kept identical across clips. Table 1 details the Dependent Variables: manipulations, including the footage used. Politicians’ Opinions The voiceover of the 32 clips, indicating the experimental manipulation, was as follows: Immediately after watching each clip, politicians were asked six questions regarding their [Issue (identical in all clips): Today, in personal beliefs and planned behavior with Brussels, a demonstration was held for bet- regard to the protest they had just watched. ter treatment of undocumented migrants. Politicians rated the following three state- The demonstrators want a more humane ments on 0 to 10 scales (totally disagree to asylum policy and clearer regularization totally agree): “The demonstrators in this clip criteria. Currently, the procedure is too arbi- attracted my interest” (interest); “I sympa- trary and takes too much time. As a result, thize with the demonstrators in this clip” asylum-seekers often stay in a state of (sympathy); and “I agree with the demonstra- uncertainty for months. The procedure tors in this clip” (agree). Representatives needs to become shorter and clearer.] [Num- were also given the following three questions bers: There were about 500 participants// to be answered on 0 to 10 scales (definitely more than 5,000 participants which was not to definitely yes): “Would you consider much less//more than expected.][Diversity: referring to this demonstration in an informal Mainly undocumented migrants took part in conversation with colleagues?” (conversa- the march, with groups of Afghan and Afri- tion); “Would you consider taking a public can asylum-seekers at the head of the stance as a result of this demonstration (tweet, demonstration.//A wide variety of associa- opinion piece, interview,…)?” (stance); and tions took part in the march. Sympathizing “Would you consider taking formal political Belgians participated next to the asylum- action as a result of this demonstration (for seekers themselves.] [Unity: Divergent example, asking an oral or written question in voices could be heard in the demonstration. parliament)?” (formal). Some demanded a clearer asylum policy, The six measures were designed to tap the others were demonstrating for peace in salience, position, and intended action beliefs Syria. Everybody seemed to have their own of the representatives, each belief being cov- reason for taking part.//Everybody took to ered by two direct measures. Interest and con- the street for the same reason, with one versation tap the salience belief (Cronbach’s Wouters and Walgrave 371

Table 1. Overview of Vignette Manipulations

Variable Aspect Low High

Worthiness text Grim atmosphere, shop-window Serene atmosphere, no incidents, broken, arrests, disruption initiated calm and peaceful demonstrators. by protestors. visual Image of broken window, demonstra- Demonstrators standing and walking tors fighting police, police arresting calmly, child on shoulders, child demonstrators. walking.

Unity text Diverging voices, asylum policy and Single message, same reason to peace in Syria, various reasons to protest, repetition of asylum policy protest. demand. visual Banner showing text about war. Only asylum policy banners.

Numbers text About 500, less than expected. More than 5,000, more than expected. visual No air shot, images showing empty Air and pan shots showing big spots in the demonstration. crowds, images full of people.

Commitment text Future unsure, no further action Second action in a row, next protest planned. planned. visual Participant drinking and smoking, Participant using megaphone to chant, standing still, hands in pockets. participant shaking clenched fist.

Diversity text Only Afghan and African asylum- A wide variety of associations; Bel- seekers participate. gians next to asylum-seekers. visual Images of non-white participants Images of white participants as well. only.

Alpha = .81), sympathy and agree the posi- of the protest affect elected officials’ opinion tional belief (Cronbach’s Alpha = .87), and formation on top of their predispositions. stance and formal relate to intended action We measured politicians’ pretreatment (Cronbach’s Alpha = .81). We use averaged asylum attitudes using four questions (0 to 10 indexes of these three pairs of two questions as scale): “Immigrants must be able to preserve the dependent variables of our study. All analy- their own culture,” “Belgium must close its ses reported here were carried out separately borders to asylum-seekers,” “Illegal immi- on all six measures; the results are identical to grants must be expelled, even when they are the ones we report. integrated,” and “Asylum-seekers with chil- dren should not be placed in detention cent- ers.” These four questions form a single Elected Officials’ Pretreatment asylum stance scale (Cronbach’s Alpha = Attitudes .69). Second, we expect left-wing party politi- We questioned politicians’ pretreatment atti- cians to think systematically more positively tudes about the asylum issue. We measured about the protests; we thus include a left- these attitudes at the beginning of the survey, party dummy in our statistical models. Third, about 25 minutes before respondents were the regional affiliation of a politician may be exposed to the clips. These attitudes can be relevant. Asylum and migration is a more considered a measure of the degree to which important and contentious issue in the Dutch- the elected officials held beliefs congruent speaking (Flanders) than in the French-speaking with the protest signal prior to their exposure (Wallonia and Brussels) part of the country. to the protest. The question is whether features Flanders has seen the rise of one of the 372 American Sociological Review 82(2) strongest right-wing anti-immigration parties officials’ beliefs—the extent to which they esti- in Europe (Vlaams Belang), whereas Wal- mated the protest issue to be important (sali- lonia never had a successful anti-immigration ence), the degree to which they substantially party. Therefore, Flemish and Walloon politi- agreed with the protesters (position), and cians should think and act differently when it whether they intended to take action following comes to asylum protest, with representatives up on the protest (action). The models control from the Flemish region likely to be more for elected representatives’ pretreatment asylum skeptical about immigration in general and beliefs and for aspects of the design (clip order). asylum-seekers’ claims in particular. Finally, It is no surprise to find that effects of we expect politicians who specialize in asy- elected officials’ prior beliefs are substantial. lum matters will consider the issue as more What politicians think about immigration and salient and will be more likely to consider asylum-seekers before exposure strongly acting on the protest signal. A dummy gauges determines their opinions. Due to their struc- whether a politician, when asked in the sur- tural (region, party, specialization) and ideo- vey about her most important working issues, logical (asylum stance and party) positions, listed asylum/immigration as one of her three elected officials’ beliefs are already, to vary- most important issues. ing extents, congruent with the claim embedded in the protest signal, irrespective of who the protesters are and how they behave. All Controls but two of the 12 coefficients tapping effects Our models take into account the viewing order of elected representatives’ predispositions in of the clips (remember that clip order was ran- the three models point toward strong and sig- domized within sets). Salience and position nificant effects. Politicians from left-wing scores vary systematically between the first and parties and those holding generally positive second clips, with politicians reporting lower asylum stances before being exposed to the salience and less congruent positions after the clips tended to find the issue more important, second clip. This might be a consequence of the agreed to a greater extent with the protest reduced novelty of the second exposure. By claim, and were more likely to consider controlling for clip order, we can gauge net undertaking action. A politician’s specializa- effects of the variables of interest. tion in asylum/immigration and belonging to the Dutch or French regional group also pro- duced salience and action effects. Our results Results suggest that these predispositions matter a Because the observations in our dataset are great deal. not independent, we present results in the We calculated predicted values based on the form of random-intercept linear multilevel models in Table 2 for the most “extreme” models. Each elected representative rated two elected representatives. A French-speaking pol- TV clips; clips (level 1) are therefore the unit itician from a left-wing party, who specialized of analysis and are modeled as nested in poli- in asylum and held pretreatment stances fully in ticians (level 2). Multilevel modeling reflects favor of asylum-seekers, rated the protest as the hierarchical structure of our data, controls 9.37 on the 0 to 10 salience scale. The exact for intra-rater correlation (Rossi and Nock mirror image of this politician scored the pro- 1982), is frequently used, and is considered test merely at 2.99 on the same scale. We the preferred strategy for analyzing vignette obtained similar results for position (8.59 ver- studies (Hox, Kreft, and Hermkens 1991; sus 2.39) and action (9.54 versus 2.21) effects. Wallander 2009). We used the xtmixed com- The model diagnostics at the bottom of the mand in Stata to fit the models and the mar- table add weight to the conclusion that preexist- gins command to calculate predicted values. ing beliefs matter strongly for opinion forma- Table 2 shows results of three models assess- tion in reaction to incoming societal signals. ing the effect of protest features on elected The variables at the politician level explain a Wouters and Walgrave 373

Table 2. Results of Random-Intercept Linear Multilevel Regressions Predicting Salience, Position, and Intended Action Beliefs of Elected Officials

Salience Position Action

Coef. SE Coef. SE. Coef. SE

Politician Features Asylum Stance .576*** .126 .865*** .105 .496*** .154 Left-Wing Party .915*** .322 .923*** .268 .942* .393 Region (Flemish) –1.002*** .278 –.410 .230 –1.243*** .338 Specialization 1.154* .470 –.110 .392 2.287*** .572

Protest Features Worthiness .032 .097 .381*** .105 .116 .111 Unity .279** .097 .353*** .105 .350** .111 Numbers .345*** .097 .282** .105 .439*** .111 Commitment .181 .097 .109 .105 .156 .111

Control Clip Order (second) –.312*** .097 –.612*** .105 –.163 .111

Constant 3.326*** .667 2.075*** .566 2.549*** .803 N Politicians 244 245 242 N Clips 486 487 478 Chi² (df) 130.41 (9)*** 253.48 (9)*** 103.07 (9)*** Log Likelihood –945.68 –922.466 –1005.33 Log Likelihood Empty –1054.55 –1072.58 –1096.62 Model Variance Elite Level 3.004 (4.245) 1.817 (3.517) 4.550 (5.917) (empty) Explained Variance Elite 29.2% 48.3% 23.1% Level Variance Clip Level 1.147 (1.298) 1.342 (1.702) 1.448 (1.598) (empty) Explained Variance Clip 11.6% 21.2% 9.4% Level

*p < .05; **p < .01; ***p < .001 (two-tailed tests).

good deal of the variance (29.2 percent, 48.3 In summary, we find evidence of strong percent, and 23.1 percent, respectively). receiver effects; representatives’ predisposi- These findings indicate that congruence tions strongly matter. Do features of the pro- between public opinion signals and elected test signal exert an effect on top of these officials’ opinions strongly depends on the potent effects of politicians’ features? In other ideology and structural position of the repre- words, can protesters—by who they are and sentative in question. Moreover, the fact that how they behave—send cues to politicians our first results match these rich “real world” that affect their opinions? Our results suggest observations about, for example, regional and they can. Figure 1 shows predicted values for party differences, suggests that the participat- all protest features based on the models in ing politicians experienced the experimental Table 2. All other variables are kept at their treatment as realistic and were properly mean; significant effects are marked with engaged with the stimuli. stars. 374 American Sociological Review 82(2)

Worthiness Unity 6.5 6.5 6.3 6.3 *** ***

e 6.1 e 6.1 5.9 5.9 ** 5.7 5.7 -10 Scal -10 Scal 5.5 5.5

on 0 5.3 on 0 5.3 5.1 5.1 **

Rang 4.9 Rang 4.9 4.7 4.7 4.5 4.5 Salience PosionAcon Salience PosionAcon

Low High Low High

Numbers Commitment 6.5 6.5 6.3 6.3

e 6.1 ** e 6.1 5.9 *** 5.9 5.7 5.7 -10 Scal -10 Scal 5.5 5.5

on 0 5.3 on 0 5.3 5.1 *** 5.1

Rang 4.9 Rang 4.9 4.7 4.7 4.5 4.5 Salience PosionAcon Salience PosionAcon

Low High LowHigh

Figure 1. Predicted Values of Salience, Position, and Intended Action Effects by Protest Feature *p < .05; **p < .01; ***p < .001 (two-tailed tests).

Whether demonstrators’ behavior in the formation is to come across as united and march is calm and serene or turns (mildly) vio- numerous. High numbers alter politicians’ lent produces a clear position effect. Demon- calculations and generate effects on all three strators behaving in a worthy manner convinced relevant beliefs (salience, position, and representatives to agree or sympathize with action). Effect sizes are modest, again, and their claim—the effect size is .4 on a 0 to 10 range from .3 (position) to .4 (action). That scale. Displays of worthiness create sympathy power is in numbers is a more or less estab- and goodwill. Worthiness does not bear any lished (but rarely explicitly empirically scru- significant effect on salience or action. The tinized) fact in social movement theory, and sheer fact that we find an effect of worthiness on we see it confirmed here for the first time, as position is a solid finding; politicians’ policy far as we know, in an experimental fashion positions on issues are supposed to be stable controlling for possible confounding factors. because they are grounded in their party’s offi- Unity is also effective for influencing cial stance and in the ideology to which they elected officials’ beliefs. The conditions with publicly adhere, which we showed earlier to be unified demonstrators received higher sali- the case. Moreover, we fully control for respond- ence, position, and action ratings, with effect ents’ pretreatment position on the issue by a sizes of .27 for salience, .36 for position, and battery of questions incorporated in the model. .35 for action. In the low unity condition, The crux of staging a protest that success- demonstrators were presented as having dis- fully affects elected officials’ opinion parate claims, ranging from specific claims Wouters and Walgrave 375 related to asylum-seekers to general claims unity, numbers, and commitment—scores, on about peace in Syria. Broadcasting messier, average, 6.4 on the position scale; protest that less coherent claims appears to be detrimental scores low on all factors comes out at 5.3. The to the strength of the protest signal, even effects for action (5.3 versus 4.2) are similar, when the messy part of the claim refers to a and for salience they are a little smaller (6.1 position no elected official would disagree versus 5.3). Moreover, the models in Table 2 with (peace in Syria). If the impression exists indicate that some of the variance at the news- that not all participants are on the same page, item level is successfully explained by the elected officials perceive protestor grievances manipulated protest features. This holds espe- as less salient, are less apt to agree with the cially for the explained variance of the posi- demonstrators, and are less prepared to take tion model (21.4 percent explained variance action. Having a focused, coherent, and non- at the clip level), and to a lesser extent for the ambivalent claim clearly matters. Moreover, salience (11.6 percent) and intended action because peace in Syria (the cue that intro- (9.4 percent) models. duced the low unity trait of the demonstra- The comparison between features of the tion) is regarded as a consensus issue in signal and the predispositions of elected offi- Belgian politics, we can convincingly claim that cials puts the results further in perspective. the coherence of the message, not disagreement Clearly, the WUNC effects we find are more with the additional claim, causes the position modest compared to the sizable effects of effect. elected officials’ pretreatment beliefs and ide- Committed demonstrators, finally, add ology. Who is at the receiving end of the weight to the saliency beliefs of MPs, although protest signal obviously matters most: politi- the effect is only marginally significant ( p = cians are professional opinion-holders with .063). If we disaggregate the saliency scale in crystallized arguments and strong world- its two component variables—informally talk- views. This holds for a sensitive and conten- ing with colleagues and explicitly assessing tious issue such as asylum. Our pick of the the importance of the protest issue—and run asylum issue probably resulted in a conserva- separate models (not reported in the table or tive test: protests dealing with newer issues figure), we see that commitment has a signifi- likely have more leeway in affecting elected cant effect on informal talk ( p = .014) but not officials’ opinion formation. Yet, the mere on direct salience ratings ( p = .397). So, our fact that we find these protest effects while evidence produces a modest indication that simultaneously controlling for the strong pre- demonstrators who manage to show that they dispositions of elected officials adds to our are not likely to give up convince political expectation that politicians are indeed respon- representatives that they deserve attention. By sive to incoming societal signals, and that displaying that their protest is here to stay, such signals can shape how elected officials protestors succeed in being the subject of con- think about political issues. versation among elected officials. Our finding that both preexisting beliefs All told, features of a protest signal do and protest features in the clips matter for matter for elected officials’ opinion forma- post-exposure opinions raises the question of tion. A clear pattern emerges: whereas high whether elected representatives are differently unity and high numbers are consequential for affected by protest features depending on all aspects of elected officials’ beliefs, high their predispositions. We tested whether the worthiness results in position effects and protest feature effects are found across-the- commitment adds some (modest) weight to board, or rather, whether they are moderated the perceived saliency of demonstrators’ by representatives’ predispositions. For each claims. The aggregate effect of all WUNC of the three outcome variables, we ran models elements together is quite substantial. Protest interacting all four protest features with each that maximizes WUNC—high worthiness, of the four features of representatives (results 376 American Sociological Review 82(2) reported in Part D of the online supplement). theorize about what it is in societal signals Of these 48 interaction coefficients (3 × 4 × 4) that matters for representatives. The worthi- none were significant; no interaction coeffi- ness of protesters sends a signal about the cients passed the p = .05 significance thresh- type of people supporting a certain position, old. In summary, we do not find evidence that the unity they display elucidates the exact elected officials’ predispositions moderate the nature of this position, their numbers convey effects of protest features. Rather, it seems as obvious information about the size of the if politicians, irrespective of their existing public opinion segment supporting that posi- beliefs, assess and interpret these incoming tion, and their commitment yields a cue of cues similarly. The WUNC effects on politi- how salient the issue is to them. cians appear to be universal, as anticipated by We put this theory to the test drawing on Tilly. WUNC seems to mobilize the most an original experimental design. We gained basic democratic responsiveness among rep- access to large numbers of actual national and resentatives; respondents were, for example, regional elected representatives and exposed almost all affected by the larger size of a this hard-to-reach but crucial group to manip- protest, probably simply because numbers are ulated television news items of a protest important in the calculations of any politician march dealing with the asylum issue. We interested in re-election. found that who the protestors are and how they behave influences elected representatives’ opinion formation. In particular, pro- Conclusions tests that mobilize demonstrators who agree What elected representatives think about among themselves, share a single claim, and issues is important. It affects what they under- bring many people to the streets impress take in the political arena and which initiatives elected officials. Unity and numbers signifi- they support. While we have some knowledge cantly change representatives’ opinions regard- about where elected officials look for informa- ing the importance of the underlying issue. tion about public opinion and how they form a These cues even alter representatives’ posi- perception of what the public wants, little to tions and make them willing to undertake no work explores whether and how their own action. When demonstrators appear worthy, opinions about issues are influenced by these elected officials’ issue positions are updated. perceptions. We argued that clear and compel- In addition, protesters revealing themselves ling signals of public opinion, or of a segment as committed to the cause might affect the of public opinion that is relevant to a politi- perceived importance of the issue. cian, can influence representatives’ own opin- These effects of the features of a protest ions, because acting in line with what the signal come on top of strong and pervasive public, or a specific segment of the public, effects of elected officials’ predispositions. wants is the most likely route to re-election (or What politicians think or believe before expo- so elected officials think). Politicians want to sure to a societal signal strongly affects their know how many people support a certain opinion formation on that issue. This should issue position, they look for cues about what come as no surprise: representatives are in the these people want and who these people are, business of having opinions on societal mat- and they search for information on how much ters, so their views should be crystallized and these people care about getting what they strong. Our findings show, however, that want. We held that the signal sent by protest- elected officials’ opinions are not entirely ing citizens may contain all these bits of infor- stable. Elected officials do update their opin- mation and may thus affect what representatives ions in reaction to external stimuli; the opin- think about political issues. ions they hold are affected by features of More concretely, we drew on Tilly’s influ- signals coming in from society. Interestingly, ential but untested WUNC account to further our results show that elected officials’ Wouters and Walgrave 377 predispositions do not function as a filtering could be different. Under a more authoritarian screen. We found no evidence that some regime, for example, more aggressive and elected officials are more responsive than disruptive protests may be needed to bring the others to some protest features. Rather, message across. WUNC features similarly affected politicians Our study contributes not only to the gen- from the left and the right, specialists and eral field of politicians’ opinion formation, non-specialists, and those who held pro-asylum but also to the more specialized field of social stances and those who did not. This finding movements and social movement outcomes. hints that Tilly’s WUNC framework, tapping The presence of elite allies is considered a into the idea of popular sovereignty as the crucial element for social movement success. origin of power in democracies, may indeed Whereas most studies of protest effects treat be universal and affect all politicians. the presence of elite allies as an exogenous Substantively, our findings specify how factor facilitating (or constraining) protest protest might be an effective weapon of the impact, our study provides evidence that pro- weak. Having other means and better access, test can shape elected officials’ opinions and, the resource rich do not need to rely on protest as such, that movements can to some extent to communicate their preferences to decision create the conditions for their own success. makers. People from weaker societal groups, Protest can push representatives toward in contrast, are probably more dependent on becoming elite allies who, in turn, are a criti- contentious forms of preference communica- cal resource in the political arena. tion. By displaying some features rather than One of the major downsides of experi- others (or making sure that journalists pick up ments is their limited external validity. We on these features), their protest signals will believe that the present study does rather well likely be better received by politicians. Our regarding both experimental and mundane experiment dealt with the issue of asylum; the realism. We exposed real politicians (not stu- demonstrators shown in the clip and men- dents) to professionally edited news items tioned by the voiceover came across mainly as (that came across as real) about a topic they being foreign and poor. In that sense, their learn about primarily through the news (pro- threat to the elected officials was not very test) in a realistic setting (their office, not a credible or impressive. Still, we found meas- laboratory). The news clips were realistically urable changes in representatives’ beliefs after designed: police archive and news media data being confronted with this relatively resource- show that asylum-seeker protests vary on the poor segment of the public. This raises the features we manipulated. The real world is, of question whether similar, maybe stronger, course, far more dynamic and complex com- effects would emerge if representatives were pared to the single snapshot we presented to confronted with a contentious event staged politicians; protest actions can inspire counter and populated by stronger societal groups, movements and trigger reactions from lobby- such as teachers or employers. ists or other elected officials, and these sig- The fact that we found worthiness to have nals, in turn, might reach and influence a positive effect on politicians’ own opinions elected officials and counterbalance the pro- about asylum might be attributed to the fact test signal. The great advantage of an experi- that resource-poor individuals are required to mental approach is precisely that it comes display their deservingness, and that they can with full researcher control, is unrivaled when do so by behaving peacefully. Previous work it comes to pinpointing causality, and allows on the effect of disruptiveness led to mixed one to isolate the effects of protest feature results, but we found worthiness to exert a factors, cancelling out the noise that is omni- positive effect. It is not unthinkable that for present in the real world. We tested and care- other issues and other groups, or in other con- fully disentangled the effects of four relevant texts, the effect of worthiness/disruptiveness features of protest and found that they matter 378 American Sociological Review 82(2) differently for three different relevant opin- Our point is that the experimental approach ions of a large group of real elected politi- put forward in this study has many potential cians. Such a level of detail and thoroughness extensions, and such research can speak to in establishing the exact mechanism of how puzzles that are at best difficult to address protest features may lead to political out- with prevailing observational methods. comes is unusual. We believe the theoretical Experiments have been on the rise in political advances that can be made with experimenta- science (Druckman et al. 2006; Iyengar 2011), tion largely outweigh its downsides. One may but far less so in sociology (Jackson and Cox challenge the generalizability of what we 2013). We believe our study shows that found, of course; we base our conclusions on experimentation with elected officials is fea- data from one country for one issue, and for sible and produces findings that can help us one possible manipulation of WUNC. In fact, disentangle complex causal processes that are our study begs for replication on other issues, very hard to tackle in other ways. countries, and different manipulations to further refine, or contradict, what we found. We believe that our novel experimental Appendix approach to representatives’ opinion forma- Part A. Dependent Variables tion opens up avenues for further work on Question Wording how politicians process and follow-up on public opinion signals more generally. Protest We will now show you two television news is not the only public opinion signal reaching items. These items are fictional and both politicians, and it would be valuable to apply deal with a demonstration on exactly the similar experimental designs to signals com- same topic. Both demonstrations developed ing from interest groups, for example. Maybe differently and, therefore, the journalist other groups’ signals are processed differently made different news items. We would like and, instead of cues about the size of the you to watch each news item carefully. population and their unity, perseverance in Imagine that the news items are real, and lobbying, their financial means, or the useful- that the protest took place as shown in the ness of the policy information they can pro- item. After each item we will ask you several vide might matter. Another extension would questions on how you personally experi- be to examine to what extent politicians’ enced the specific protest, as depicted in the opinion formation differs from how ordinary news item. If you are answering these ques- citizens form their opinions in reaction to tions on your own computer, please make protest or other public opinion signals. This sure that the audio is on and that you can could tell us whether politicians are unique or listen to the news item. Click on ‘next’ to react in similar ways as everyday citizens. start the news clip. Wouters and Walgrave 379

[Separate page with video clip: play clip] [Separate page with questions] We are interested in your personal opinion about the demonstration as shown in this first/ second news clip. Indicate on a scale from 0 to 10 to what extent you agree with the following statements.

Totally not agree Totally agree

0 1 2 3 4 5 6 7 8 9 10

1. The demonstrators in this clip attracted my interest. 2. I sympathize with the demonstrators in this clip 3. I agree with the demonstrators in this clip.

Besides your personal opinion, we are also interested in potential actions you yourself could possibly take as a result of the demonstration in this first/second news clip. On a scale of 0 to 10, to what extent…

Definitely Not Definitely Yes

0 1 2 3 4 5 6 7 8 9 10

4. … would you consider referring to this demonstration in an informal conversation with colleagues? 5. … would you consider taking a public stance as a result of this demonstration (tweet, opinion piece, interview,…)? 6. … would you consider taking formal political action as a result of this demonstration (for example, asking an oral or written question in parliament)? 380 American Sociological Review 82(2)

Part B. Descriptives of Dependent Variables

Acknowledgments N° 12M6415N). Stefaan Walgrave gratefully acknowledges support of the European Research Council For helpful feedback, we thank Shanto Iyengar, Sarah (Advanced Grant ‘INFOPOL’, N° 295735) and the Soule, Doug McAdam, Susan Olzak, Luzia Helfer, Bert Research Fund of the University of Antwerp (Grant N° Fraussen, Tobias Konitzer, participants at the SCANCOR 26827). Stefaan Walgrave (University of Antwerp) is the Seminar and the Political Sociology workshop at Stan- principal investigator of the INFOPOL project, which has ford University, and the ASR editors and reviewers. For additional teams in Israel (led by Tamir Sheafer) and video clip creation, we thank Caroline Van Ranst, Mela- Canada (led by Stuart Soroka and Peter Loewen). The nie van den Plas, Jules Debrock, Jan Van Delm, and protest impact study was conducted only in Belgium. Johanne Montay.

Funding Note Ruud Wouters gratefully acknowledges support of the 1. The data we used in this study are available for Research Foundation Flanders – FWO (Post doc Grant replication and validation. A dataset containing the Wouters and Walgrave 381

variables we used here, as well as the coding syn- and What It Gets. New York: Cambridge University tax and the statistical commands, is available in the Press. online supplement. Burstein, Paul, and April Linton. 2002. “The Impact of Political Parties, Interest Groups, and Social Move- ment Organizations on Public Policy: Some Recent References Evidence and Theoretical Concerns.” Social Forces Amenta, Edwin, Neal Caren, Elizabeth Chiarello, and 81(2):380–408. Yang Su. 2010. “The Political Consequences of Carnes, Nicholas. 2013. White-Collar Government: The Social Movements.” Annual Review of Sociology Hidden Role of Class in Economic Policy Making. 36:287–307. Chicago: University of Chicago Press. Amenta, Edwin, Neal Caren, and Sheera Joy Olasky. Cress, Daniel M., and David A. Snow. 2000. “The 2005. “Age for Leisure? Political Mediation and the Outcomes of Homeless Mobilization: The Influ- Impact of the Pension Movement on U.S. Old-Age ence of Organization, Disruption, Political Media- Policy.” American Sociological Review 70(3):516–38. tion, and Framing.” American Journal of Sociology Amenta, Edwin, Bruce G. Carruthers, and Yvonne 105(4):1063–1104. Zylan. 1992. “A Hero for the Aged? The Townsend Dahl, Robert. 1973. Polyarchy: Participation and Oppo- Movement, the Political Mediation Model, and U.S. sition. New Haven, CT: Yale University Press. Old-Age Policy, 1934–1950.” American Journal of Dalton, Russell J. 2008. “Citizenship Norms and the Sociology 98(2):308–339. Expansion of Political Participation.” Political Stud- Andrews, Kenneth T. 2004. Freedom Is a Constant Strug- ies 56(1):76–98. gle: The Mississippi Civil Rights Movement and Its della Porta, Donatella, and Mario Diani. 1999. “Forms, Legacy. Chicago: University of Chicago Press. Repertoires and Cycles of Protest.” Pp. 165–92 in Arnold, R. Douglas. 1990. The Logic of Congressional Social Movements: An Introduction, edited by D. Action. New Haven, CT: Yale University Press. della Porta and M. Diani. Oxford, UK: Blackwell. Atzmüller, Christiane, and Peter M. Steiner. 2010. De Nardo, James. 1985. Power in Numbers: The Politi- “Experimental Vignette Studies in Survey Research.” cal Strategy of Protest and Rebellion. Princeton, NJ: Methodology 6(3):128–38. Princeton University Press. Bartels, Larry M. 2008. Unequal Democracy: The Politi- Deschouwer, Kris, and Sam Depauw. 2014. Representing cal Economy of the New Gilded Age. Princeton, NJ: the People: A Survey among Members of Statewide Princeton University Press. and Substate Parliaments. Oxford, UK: Oxford Uni- Baumgartner, Frank R., and Bryan D. Jones. 1993. Agen- versity Press. das and Instability in American Politics. Chicago: Downs, Anthony. 1957. “An Economic Theory of Politi- University of Chicago Press. cal Action in a Democracy.” Journal of Political Belchior, Ana Maria. 2014. “Explaining MPs’ Percep- Economy 65(2):135–50. tions of Voters’ Positions in a Party-Mediated Rep- Druckman, James N., Donald P. Green, James H. Kuk- resentation System: Evidence from the Portuguese linski, and Arthur Lupia. 2006. “The Growth and Case.” Party Politics 20(3):403–415. Development of Experimental Research in Politi- Benford, Robert D., and David A. Snow. 2000. “Framing cal Science.” American Political Science Review Processes and Social Movements: An Overview and 100(4):627–35. Assessment.” Annual Review of Sociology 26:611– Erikson, Robert S., Michael B. MacKuen, and James A. 39. Stimson. 2002. The Macro Polity. New York: Cam- Biggs, Michael, and Kenneth T. Andrews. 2015. “Protest bridge University Press. Campaigns and Movement Success: Desegregating Etzioni, Amitai. 1970. Demonstration Democracy. New the U.S. South in the Early 1960s.” American Socio- York: Gordon and Breach. logical Review 80(2):416–43. Fassiotto, Magali, and Sarah Soule. 2015. “Loud and Blumler, Jay G., and Dennis Kavanagh. 1999. “The Third Clear: The Effect of Protest Signals on Congressional Age of Political Communication: Influences and Fea- Attention.” Unpublished paper, Stanford University, tures.” Political Communication 16(3):209–230. Stanford, CA. Broockman, David E., and Christopher Skovron. 2013. Fenno, Richard F. 1978. Home Style: House Members in “What Politicians Believe about Their Constituents: Their Districts. Boston: Little Brown. Asymmetric Misperceptions and Prospects for Con- Ferree, Myra Marx, William Gamson, Jurgen Gerhards, stituency Control.” Unpublished paper, University of and Dieter Rucht. 2002. Shaping Abortion Discourse: California, Berkeley. Democracy and the Public Sphere in Germany and the Burstein, Paul. 1999. “Social Movements and Public United States. New York: Cambridge University Press. Policy.” Pp. 3–21 in How Social Movements Matter, Fishbein, Martin, and Icek Ajzen. 2011. Predicting and edited by M. Giugni, D. McAdam, and C. Tilly. Min- Changing Behavior: The Reasoned Action Approach. neapolis: University of Minnesota Press. New York: Taylor & Francis. Burstein, Paul. 2014. American Public Opinion, Advo- Gamson, William. 1990. The Strategy of Social Protest. cacy, and Policy in Congress: What the Public Wants Belmont, CA: Wadsworth. 382 American Sociological Review 82(2)

Gamson, William. 2004. “Bystanders, Public Opinion and Jones, Bryan D., Gramae Boushey, and Samuel Work- the Media.” Pp. 242–62 in Blackwell Companion to man. 2006. “Behavioral Rationality and the Policy Social Movement Studies, edited by D. Snow, S. Soule, Processes: Toward a New Model of Organizational and H. Kriesi. Oxford, UK: Blackwell Publishing. Information Processing.” Pp. 49–74 in Handbook of Geer, John Gray. 1996. From Tea Leaves to Opinion Public Policy, edited by B. G. Peters and J. Pierre. Polls: A Theory of Democratic Leadership. New London, UK: Sage. York: Columbia University Press. Key, Valdimer Orlando. 1961. Public Opinion and Amer- Gilens, Martin. 2005. “Inequality and Democratic ican Democracy. New York: Alfred A. Knopf. Responsiveness.” Public Opinion Quarterly 69(5): King, Brayden G., Keith G. Bentele, and Sarah A. Soule. 778–96. 2007. “Protest and Policymaking: Explaining Fluc- Gillion, Daniel Q. 2012. “Protest and Congressional tuation in Congressional Attention to Rights Issues, Behavior: Assessing Racial and Ethnic Minority Pro- 1960–1986.” Social Forces 86(1):137–63. tests in the District.” Journal of Politics 74(4):950–62. Koopmans, Ruud. 2004. “Movements and Media: Selec- Gillion, Daniel Q. 2013. The Political Power of Protest: tion Processes and Evolutionary Dynamics in the Minority Activism and Shifts in Public Policy. New Public Sphere.” Theory and Society 33(3–4):367–91. York: Cambridge University Press. Lang, Kurt, and Gladys Engel Lang. 2008. “Advocacy: Giugni, Marco. 1999. “How Social Movements Matter: Alternative Expressions of Public Opinion.” Pp. Past Research, Present Problems, Future Develop- 41–52 in The Sage Handbook of Public Opinion ments.” Pp. xiii–xxxiii in How Social Movements Research, edited by W. Donsbach and M. W. Trau- Matter, edited by M. Giugni, D. McAdam, and gott. London, UK: Sage. C. Tilly. Minneapolis: University of Minnesota Press. Lipsky, Michael. 1968. “Protest as a Political Resource.” Giugni, Marco. 2007. “Useless Protest? A Time-Series American Political Science Review 62(4):1144–58. Analysis of the Policy Outcomes of Ecology, Anti- Lohmann, Suzanne. 1993. “A Signaling Model of Infor- nuclear, and Peace Movements in the United States, mative and Manipulative Political Action.” American 1977–1995.” Mobilization: An International Quar- Political Science Review 87(2):319–33. terly 12(1):53–77. Luders, Joseph. 2006. “The Economics of Movement Suc- Herbst, Susan. 1998. Reading Public Opinion: How cess: Business Responses to Civil Rights Mobiliza- Political Actors View the Democratic Process. New tion.” American Journal of Sociology 111(4):963–98. York: Cambridge University Press. Mansbridge, Jane. 1999. “Should Blacks Represent Hilgartner, Stephen, and Charles L. Bosk. 1988. “The Blacks and Women Represent Women? A Contingent Rise and Fall of Social Problems: A Public Arenas ‘Yes.’” Journal of Politics 61(3):628–57. Model.” American Journal of Sociology 94(1):53–78. Mayhew, David R. 1974. Congress: The Electoral Con- Hox, Joop J., Ita G. G. Kreft, and Piet L. J. Hermkens. nection. New Haven, CT: Yale University Press. 1991. “The Analysis of Factorial Surveys.” Sociolog- McAdam, Doug. 1983. “Tactical Innovation and the ical Methods & Research 19(4):493–510. Pace of Insurgency.” American Sociological Review Hsu, Greta, Michael T. Hannan, and Özgecan Koçak. 48(6):735–54. 2009. “Multiple Category Memberships in Markets: McAdam, Doug. 1996. “The Framing Function of Move- An Integrative Theory and Two Empirical Tests.” ment Tactics: Strategic Dramaturgy in the American American Sociological Review 74(1):150–69. Civil Rights Movement.” Pp. 338–55 in Compara- Hunt, Scott A., and Robert D. Benford. 2004. “Collective tive Perspectives on Social Movements: Political Identity, Solidarity, and Commitment.” Pp. 433–57 Opportunities, Mobilizing Structures, and Cultural in The Blackwell Companion to Social Movements, Framings, edited by D. McAdam, J. McCarthy, and edited by D. Snow, S. Soule, and H. Kriesi. Oxford, M. Zald. Cambridge, UK: Cambridge University Press. UK: Blackwell Publishing. McAdam, Doug, and Yang Su. 2002. “The War at Home: Iyengar, Shanto. 2011. “Laboratory Experiments in Antiwar Protests and Congressional Voting, 1965 to Political Science.” Pp. 73–88 in Handbook of Experi- 1973.” American Sociological Review 67(5):696– mental Political Science, edited by J. Druckman, 721. D. Green, J. Kuklinski, and A. Lupia. Cambridge, McGraw, Kathleen M. 1990. “Avoiding Blame: An UK: Cambridge University Press. Experimental Investigation of Political Excuses and Jackson, Michelle, and David R. Cox. 2013. “The Prin- Justifications.” British Journal of Political Science ciples of Experimental Design and Their Application 20(1):119–31. in Sociology.” Annual Review of Sociology 39:27–49. Meyer, David S., and Sidney G. Tarrow. 1998. The Social Johnson, Erik W., Jon Agnone, and John D. McCarthy. Movement Society: Contentious Politics for a New 2010. “Movement Organizations, Synergistic Tactics Century. Lanham, MD: Rowman & Littlefield. and Environmental Public Policy.” Social Forces Miler, Kristina C. 2007. “The View from the Hill: Legis- 88(5):2267–92. lative Perceptions of the District.” Legislative Studies Jones, Bryan D., and Frank R. Baumgartner. 2005. The Quarterly 32(4):597–628. Politics of Attention: How Government Prioritizes Miler, Kristina C. 2009. “The Limitations of Heuristics for Problems. Chicago: University of Chicago Press. Political Elites.” Political Psychology 30(6):863–94. Wouters and Walgrave 383

Miller, Wakken E., and Donald E. Stokes. 1963. “Con- Tilly, Charles. 1995. Popular Contention in Great Brit- stituency Influence in Congress.” American Political ain, 1758–1834. Cambridge, MA: Harvard Univer- Science Review 57(1):45–56. sity Press Mondak, Jeffery J. 1995. “Competence, Integrity, and Tilly, Charles. 1999. “From Interactions to Outcomes in the Electoral Success of Congressional Incumbents.” Social Movements.” Pp. 253–70 in How Social Move- Journal of Politics 57(4):1043–69. ments Matter, edited by M. Giugni, D. McAdam, and Norris, Pippa, Stefaan Walgrave, and Peter Van Aelst. C. Tilly. Minneapolis: University of Minnesota Press. 2005. “Who Demonstrates? Antistate Rebels, Con- Tilly, Charles. 2003. The Politics of Collective Violence. ventional Participants, or Everyone?” Comparative New York: Cambridge University Press. Politics 37(2):189–205. Tilly, Charles. 2004. Social Movements, 1768–2004. Olzak, Susan, and Sarah A. Soule. 2009. “Cross-Cutting London, UK: Paradigm Publishers. Influences of Environmental Protest and Legisla- Tilly, Charles. 2006. “WUNC.” Pp. 289–306 in Crowds, tion.” Social Forces 88(1):201–225. edited by J. T. Schnapp and M. Tiews. Stanford, CA: Page, Benjamin I., and Robert Y. Shapiro. 1992. The Ratio- Stanford University Press. nal Public: Fifty Years of Trends in Americans’ Policy Tilly, Charles. 2008. Contentious Performances. New Preferences. Chicago: University of Chicago Press. York: Cambridge University Press. Piven, Frances Fox, and Richard Cloward. 1993. Regu- Tilly, Charles. 2010. Regimes and Repertoires. Chicago: lating the Poor: The Functions of Public Welfare. University of Chicago Press. New York: Vintage Books. Uslaner, Eric M. 1999. The Movers and the Shirkers: Popkin, Samuel L. 1991. The Reasoning Voter: Commu- Representatives and Ideologues in the Senate. Ann nication and Persuasion in Presidential Campaigns. Arbor: University of Michigan Press. Chicago: University Of Chicago Press. Walgrave, Stefaan, and Rens Vliegenthart. 2012. “The Rossi, Peter H., and Steven L. Nock. 1982. Measuring Complex Agenda-Setting Power of Protest: Dem- Social Judgments: The Factorial Survey Approach. onstrations, Media, Parliament, Government, and London, UK: Sage. Legislation in Belgium, 1993–2000.” Mobilization Schlozman, Kay Lehman, Sidney Verba, and Henry E. 17(2):129–56. Brady. 2012. The Unheavenly Chorus: Unequal Politi- Wallander, Lisa. 2009. “25 Years of Factorial Surveys cal Voice and the Broken Promise of American Democ- in Sociology: A Review.” Social Science Research racy. Princeton, NJ: Princeton University Press. 38(3):505–520. Schneider, Anne L., and Helen M. Ingram. 2005. Deserv- Weaver, Kent R. 1986. “The Politics of Blame Avoid- ing and Entitled: Social Constructions and Public ance.” Journal of Public Policy 6(4):371–98. Policy. Albany: SUNY Press. Wlezien, Christopher. 2004. “Patterns of Representation: Simon, Herbert A. 1962. “The Architecture of Complex- Dynamics of Public Preferences and Policy.” Journal ity.” Proceedings of the American Philosophical of Politics 66(1):1–24. Society 106(6):467–82. Wouters, Ruud. 2013. “From the Street to the Screen: Char- Skrentny, John D. 2006. “Policy-Elite Perceptions and acteristics of Protest Events as Determinants of Televi- Social Movement Success: Understanding Variations sion News Coverage.” Mobilization 18(1):83–104. in Group Inclusion in Affirmative Action.” American Zaller, John. 1992. The Nature and Origins of Mass Opin- Journal of Sociology 111(6):1762–1815. ion. New York: Cambridge University Press. Soroka, Stuart N., and Christopher Wlezien. 2010. Zaller, John. 1997. “A Model of Communication Effects Degrees of Democracy: Politics, Public Opinion, and at the Outbreak of the Gulf War.” Pp. 296–311 in Policy. New York: Cambridge University Press. Do the Media Govern, edited by S. Iyengar and Soule, Sarah A., and Susan Olzak. 2004. “When Do R. Reeves. London, UK: Sage Publications. Movements Matter? The Politics of Contingency and the Equal Rights Amendment.” American Sociologi- cal Review 69(4):473–97. Ruud Wouters is a post-doctoral researcher of the Strøm, Kaare. 1997. “Rules, Reasons and Routines: Leg- Research Foundation Flanders (FWO) at the Media, islative Roles in Parliamentary Democracies.” Jour- Movements and Politics research unit of the University nal of Legislative Studies 3(1):155–74. of Antwerp. His research interests are media coverage of Strömbäck, Jesper. 2008. “Four Phases of Mediatization: social movements, protest participation, and protest An Analysis of the Mediatization of Politics.” Inter- impact. national Journal of Press/Politics 13(3):228–46. Tarrow, Sidney. 1994. Power in Movement: Social Move- Stefaan Walgrave is a professor of political science and ments and Contentious Politics. Cambridge, UK: head of the Media, Movements and Politics research unit Cambridge University Press. at the University of Antwerp. His work deals with social Tilly, Charles. 1994. “Social Movements as Historically movements and protest, public opinion and elections, Specific Clusters of Political Performances.” Berke- media and politics, and the information-processing of ley Journal of Sociology 38:1–30. political elites.