Katerina Tertytchnaya* and Tomila Lankina** Sub-national Electoral Protests and Political Attitudes under Electoral.

Transcription

1 Katerina Tertytchnaya* and Tomila Lankina** Sub-national Electoral Protests and Political Attitudes under Electoral Authoritarianism Paper presented at the Annual Meeting and Exhibition of the American Political Science Association, Philadelphia, 1-4 September 2016 * Final-year DPhil candidate in politics, University of Oxford; and Fulbright-Schuman Predoctoral Fellow ( ), Columbia University katerina.tertytchnaya@stx.ox.ac.uk ** Associate Professor, Department of International Relations, London School of Economics and Political Science t.lankina@lse.ac.uk 1

2 Abstract Do subnational electoral protests increase popular opposition to authoritarianism? And, are the cognitive effects of protests resilient to autocrats tactics of police repression and media manipulation? Combining an original author-assembled protest event dataset with survey evidence from the electoral protests in Russia, we examine whether and how the spread of protests outside of the national capital affects public opinion on issues of national significance. We find that subnational street rallies increase perceptions of electoral injustice and generate support for the demands of the protesters. Nevertheless, the findings suggest that the cognitive effects of protests are not uniform, but rather conditioned by the use of regime-led repression against protesters and the framing of protest in state-controlled national media. 2

3 1 Introduction Do sub-national electoral protests matter for shaping popular attitudes towards the national non-democratic regime? 1 And, are the effects of sub-national protests on public opinion resilient to the national autocrats and their local clients tactics of repression and manipulation? These questions are of paramount importance to the ongoing debates about the implications of mass mobilization for authoritarian regime stability and resilience (DeNardo 1985; Lohmann 1994; Magaloni 2010; Meirowitz and Tucker 2013; Robertson 2013). Yet, the vast majority of studies of electoral protests have zoomed in on dramatic mass rallies converging on national capitals, and on rulers responses to these rallies in the form of repressive or manipulative tactics. The sub-national dimension of electoral protests that may form part of a broader wave of anti-regime mobilizations has received limited theoretical and empirical treatment in this literature. Drawing on the research into the significance of local conditions for the formation of public opinion on issues of national importance (Ansolabehere, Meredith, and Snowberg 2014; Branton et al. 2015; Gartner, Segura, and Wilkening 1997; Gartner and Segura 2000; Wallace, Zepeda-Millán, and Jones-Correa 2014), we argue that territorially dispersed protests that coincide in time with dramatic national rallies in state capitals can have powerful effects on citizens. Considering that individuals tend to pay heightened attention to events occurring within their own local and social milieu, we contend that sub-national protests generate more direct exposure to the protest movement than when citizens simply witness the distant metropolitan rallies on their television screens. Territorially-dispersed 1 We use the terms hybrid, non-democratic regimes and electoral authoritarianism interchangeably. We also use the terms protests, mass mobilization, mass demonstrations and mass unrest interchangeably. 3

4 post-electoral protests, that is, events taking place in sub-national localities outside of the national capital city, invite two important sets of questions in particular. First, how do such protests affect local public opinion on issues of national significance, namely fraud perpetrated in national elections? And, do they increase support for the protesters demands? Second, are the effects of regional protests on public opinion independent of regime interventions? How does state media coverage of protest and protest repression condition the cognitive effects of sub-national protests? We highlight two mechanisms in particular through which regional anti-regime protest events could influence public opinion in nondemocratic regimes. First, in a controlled media environment, regional rallies could have information revelation effects, generating awareness about issues like electoral misconduct that may not feature prominently in the national media, perhaps even altering public opinion on these issues. Second, protests may not only reveal information otherwise not available or distorted in state media, but they could also expose citizens to the state s repressive apparatus as when for instance protesters witness or hear about citizens from their region being caught up in a skirmish with the police during a rally. The electoral protests in Russia provide us with an excellent opportunity to analyze the implications of sub-national rallies for public opinion formation under electoral authoritarianism. In the aftermath of the 2011 State Duma (parliamentary) elections, tens of thousands of citizens took to the streets across Russia s eleven time zones. In these rallies, citizens who directly targeted the federal authorities and made allegations that the regime perpetrated electoral fraud (Greene 2013; Robertson 2013), advanced a variety of demands ranging from the release of political prisoners, to the repeat of the elections, to the overthrow of the national regime. The rallies have been widely referred to as electoral protests in that protesters coalesced around the issue of fraud perpetrated in the parliamentary elections and condemned it as a mechanism for the regime to stay in power. For our analysis, 4

5 we have assembled an original Protest-Event Dataset with data on sub-national rallies and instances of repression of protest events that took place during the protest wave. We analyze these data along with data from two public opinion surveys that were in the field during that time. The Russian Election Study (RES) survey organized by Timothy Colton and his collaborators and the survey led by Stephen White (Colton et al. 2014; White 2015) allow us to connect protest events occurring within a particular region with the political attitudes of those surveyed within that region. We are not aware of other studies that would link subnational protest event data covering the entire country with public opinion data that likewise cover a large number of Russia s regions. We find support for our main theoretical predictions. Sub-national street rallies appear to raise awareness of the national protest movement and to increase the likelihood that citizens will report higher levels of electoral unfairness in the parliamentary elections. We also find that sub-national protests to some extent encourage support for the broader antiregime agenda of the protesters. Simultaneously, however, the cognitive effects of protests are weaker among respondents who reside in regions where violence had been previously used against protesters, and among respondents who frequently watch political news on statecontrolled television. Citizens exposure to the regime s media messages manipulating information on the protest movement and to the suppression of rallies by the local authorities does allow the regime to achieve the intended effects of public opinion turning away from support for the protesters and their demands. Our paper is structured as follows. In the next section, we outline a theoretical framework that connects territorially dispersed anti-regime protests and regime strategies to public opinion on national politics. We then discuss our methods and empirical results. We are aware that some underlying preferences or grievances of citizens residing in Russia s various sub-national territories could be driving the frequency and size of protests, as well as 5

6 attitudes. We therefore begin our analysis by showing that the effects of protest on public opinion are robust to the use of an instrumental variables approach. We exploit variation in regional weather conditions on the day of the rallies as an exogenous source of variation in protest frequency and size. Having established that protests do not only reveal but also cause a change in regional public opinion, we proceed with the analysis of protest effects on perceptions of electoral falsifications and regional attitudes on the protest movement. We conclude by discussing our work s contribution to the broader theorizing into the implications of popular unrest for regime stability under electoral authoritarianism. 2 Theoretical Framework 2.1. The effects of sub-national protest on public opinion Electoral protests have received considerable attention in the literature on electoral authoritarian regimes that is, those regimes that allow regular elections, but manipulate the races to ensure significant advantages and usually victory for national leaders (Hollyer, Rosendorff, and Vreeland 2015; Magaloni 2010; Svolik 2012). Much of this scholarship has approached the topic of electoral protests from the vantage point of the political regime itself. Thus, scholars have explored the implications of protests for regime splits and disagreements; for perceptions of threats, insecurities and uncertainty that anti-regime rallies generate among power holders; and for change in policy and political institutions that the political leaders effect to accommodate protesters demands. The literature on the spread of revolutionary uprisings also points to the importance of anti-regime protest events for encouraging larger numbers of citizens to take to the streets (Kuran 1995; Lohmann 1994). It is less clear, however, to what extent the views and attitudes 6

7 of the wider citizenry may be themselves shaped by anti-regime protest events. The relative paucity of research into the public opinion implications of electoral rallies is puzzling. After all, political opposition leaders, their followers and other activists taking part in rallies seek to expand public support for their demands, as much as they seek to target the regime with specific messages both strategies are inter-related (Gamson 2004; Giugni 1998; Tarrow 2011). And, for citizens to willingly participate in ongoing anti-regime demonstrations, they must first come to share, and support, the demands of the protest movement (see for example Wallace, Zepeda-Millán, and Jones-Correa 2014). How can protests influence public opinion about the movement and encourage nonparticipants to join? We argue that this question is particularly salient in territorially-large settings where even the highly visible anti-regime mass mobilizations occurring in the national capital may nevertheless not resonate across the country in a uniform way. To begin to make sense of the implications of sub-national electoral protests for public opinion, we draw on a body of research into the cognitive effects of the 2006 immigration-related subnational rallies in the U.S. (Branton et al. 2015; Wallace, Zepeda-Millán, and Jones-Correa 2014). These works innovatively combine survey evidence from the Latino National Survey (LNS) with sub-national protest data collected during the protest movement. According to this scholarship, subnational protests related to an issue of national concern, such as immigration, can have powerful effects on people who reside in counties in which these rallies are occurring. These studies join influential scholarship that has highlighted the impact of local conditions on support for incumbents, citizen perceptions of the state of the national economy, and even opinions on international issues and military engagement abroad (Ansolabehere, Meredith, and Snowberg 2014; Healy and Lenz 2014; Segura and Gartner, 2013, 2014). To highlight the significance of the local-contextual influences on public attitude formation, Garther and Segura (2000) propose the theory of modified sociotropism. 7

8 This theory acknowledges that citizens tend to pay greater attention to issues that are in some way anchored in their own local or social milieu. [I]ndividuals are affected by societal experience with an issue, but their perceptions of that experience are shaped by available proximate information-salient information about how the issue affects those one knows and cares about the most, they write (Gartner and Segura 2000, 117). Issues of national significance are also processed with reference to how they are experienced within one s own social context. And, patterns of information diffusion about the issue will be affected depending on whether and how it is experienced locally (Gartner, Segura, and Wilkening 1997; Sigmund et al. 2010). What, then, are the precise causal mechanisms whereby sub-national protest events can influence public opinion on issues pursued by protesters? Modified sociotropism allows us to discern two possible ways of unpacking how the spatial or local effects of protest events operate. The first straightforward mechanism of how regional protests may influence public opinion is actually witnessing the event: the likelihood that one would personally observe a protest whether on their daily commute to work, or when taking one s children to a park during a weekend is greater the closer that event is geographically to a particular individual (see also Branton et al. 2015; Wallace, Zepeda-Millán, and Jones-Correa 2014). Space however may be also conceptualized in terms of a broader sense of belongingness to a particular locality or region. In line with such a conceptualization, individuals will pay attention to news about protest taking place in their neighborhood, county or region whether or not they personally witness it (Gartner, Segura, and Wilkening 1997). This is because localities may be regarded as bounded spaces within which interpretations of particular events are shaped and constructed. Under such a scenario, networks of friends and family diffuse and help process information about regional protest events occurring within a socially meaningful milieu. Whether or not a resident witnesses a protest or hears about it in the news, 8

9 his or her friends, relatives or colleagues might be participating in it or discussing it on- or offline. In analyzing patterns of participation in Ukraine s Euromaidan protests, Olga Onuch for instance found that discussions about joining the protest among networks of relatives and friends were particularly instrumental in bringing people out to the streets (Onuch 2015). Both of these conceptualizations of spatial or local effects are consistent with what we know about how the particular rallies analyzed in this paper unfolded and, more generally, about regional identities in Russia. We know for instance that during the protests, not only did activists staging events strategically organize them in popular locations like prominent public squares, but they also used loudspeakers to lure bystanders to their cause and set up tents in which ordinary people could sign up to become election monitors (Lankina and Skovoroda 2016). This strategy would increase the likelihood that bystanders would be directly exposed to protesters demands, even at random. Furthermore, researchers have found that the region (oblast, Republic, krai) that is, the administrative unit equivalent to the state in America is more than just an administrative entity, or more specifically for our analysis, just a convenient proxy for geographic proximity to particular events. John O Loughlin and his collaborators have argued: [A] region is more than an artefact, but stands for a well-developed geographic consciousness, dating from pre-soviet times, and promoted since 1989 by ethnic-national alignments and contrasting local fortunes in the economic transition (O Loughlin, Shin, and Talbot 1996). Echoing this premise, ethnographic accounts and public opinion surveys have revealed that Russians maintain strong attachments to, and identification with, their administrative region (Rogers 2005). Thus, we anticipate that citizens would be particularly alert to events of national significance in our case, protests against electoral fraud if these events could be also anchored in occurrences within their particular sub-national unit. We also propose that regionally-occurring protests will have information-revelation effects about the national 9

10 protest movement and about electoral fraud and will encourage citizens to support its demands. Explicitly stated: H1: As the number of regional protests increases, awareness of, and support for, the demands of the national protest movement also increases. 2.2 The conditioning effects of regime strategies Up to now, our discussion has proceeded with the understanding that broadly similar assumptions could be derived from analyzing the effects of subnational rallies on public opinion under various political regimes. However, the contexts in which protesters aim to generate support for their demands differ in key respects across political systems. Electoral authoritarian regimes possess a stronger propaganda advantage than do democracies; in what has become known as the protest paradigm, even in western settings the media tend to subtly stigmatize street rallies and individuals challenging the status quo (DeLuca, Lawson, and Sun 2012; McLeod and Detenber 1999; Smith et al. 2001). Electoral authoritarian regimes also far more readily engage their repressive apparatus against anti-regime protesters than would democratically accountable governments (DeNardo 1985; Kricheli, Livne, and Magaloni 2011; Lohmann 1994). For these reasons, we propose that we cannot fully understand the cognitive effects of protests without taking regime strategies into account (see also DeNardo 1985). We have already highlighted that in a controlled media environment, regional protests can increase awareness of the nationally-prominent mobilizations and perhaps sympathy for the protesters demands. Yet, the media in non-democratic states do not simply crowd out 10

11 information on protest. Recent literature on media practices in authoritarian states suggests that leaders often resort to the manipulation of political news, rather than to old-fashioned censorship techniques (Guriev and Treisman 2015). By encouraging the media to reproduce fabricated allegations against protesters or to ensure disproportionate coverage of pro-regime counter-mobilizations, authoritarian regimes aim to distort citizens incentives of joining protests and to preclude the demands of the opposition from becoming publicly known. Given that media consumption powerfully affects public opinion in democracies and nondemocracies (see for example De Boef and Kellstedt 2004; Maestas et al. 2008; Reuter and Szakonyi 2013) and that citizens often tend to regard controlled authoritarian state media as trustworthy (Mickiewicz 2006; Smyth and Oates 2015), we would expect exposure to the regime s rhetoric on protest via the mass media to condition the effects of protest on public opinion. As one strand of media theorizing reminds us, the consistency of narratives across any media environment increases the acceptance of the dominant message (Atkeson and Maestas 2014; Chong and Druckman 2007b). Thus, we anticipate that citizens who are more systematically exposed to the state s rhetoric on the protest events, that is, those who more frequently watch political news on state-controlled media, will be less likely to support the demands of the wider protest movement and to report high levels of fraud in previous elections than those who watch political news on state-controlled media less frequently. Explicitly stated: H2: The cognitive effects of regional protests will be lower among individuals who are more frequently exposed to national state-controlled media As noted above, authoritarian regimes far more readily engage their repressive apparatuses against anti-regime protesters than would be the case in democracies where such 11

12 actions would be effectively scrutinized at the ballot box. This implies that citizens may not only witness or hear about local protests, but may also experience, witness, or hear about police brutality or other forms of protest suppression in their community or region. Extending the logic of the study that found that war casualties from one s county dampen support for the national war effort (Gartner and Segura 2000), one might expect citizens to condemn repression and violence, and, by extension, extend their support for the cause of the protesters who are victims of such repression, particularly if it occurs in one s own locality. Nevertheless, there is considerable disagreement on this issue in recent theorizing into repression in autocracies. Several studies of regime responses to street unrest have in fact suggested that repression can be a double-edged sword for governments (Gehlbach, Sonin, and Svolik 2015; Kricheli, Livne, and Magaloni 2011; Wintrobe 1998). Some researchers have indicated that although suppression of the protest movement may weaken the opposition, it can also signal to the electorate that the regime is crumbling under pressure, attempting as it does to hide electoral manipulations (Schedler 2002). Others have conjectured that as the threat, or the actual occurrence, of political violence becomes more salient, risk-averse citizens may well shy away from supporting protesters perceived to be engaging in violent post-electoral battles (Magaloni 2006, 2010). Protest events in respondents regions can powerfully convey the threat posed by rallies and their highly destabilizing potential. By heightening spectators anxiety (Branton et al. 2015) and conveying the image of a country gradually regressing towards chaos, violent clashes between the regime and its opponents have the potential to undermine support for the demands of the protest movement and to dampen perceptions of electoral unfairness. These alternative sets of outcomes warrant the articulation of hypotheses that are sensitive to both of these possibilities. We therefore articulate our final two hypotheses as follows: 12

13 H3a: Citizens exposure to violent suppression of regional protests will enhance their support for the national protest movement and its demands. H3b: Citizens exposure to violent suppression of regional protests will dampen their support for the national protest movement and its demands. 3 Empirical Strategy To explore how protest events influenced public opinion, we combine two public opinion surveys that were in the field between January and May 2012 with our protest dataset. 2 Interviews for the 2012 Russian Election Study (RES) and the Stephen White surveys (Colton et al. 2014; White 2015) were conducted face-to-face across forty of Russia s regions between January and May The White Survey was conducted in January 2012 and the RES between April and May The two surveys contain information about the date of every respondent s interview, the subnational district and the region in which they reside. This allows us to match survey respondents with the number of protests that took place in their administrative region between 4 December 2011 (the day of the Duma election), and the day of the respondent s interview. Likewise, the surveys ask respondents a range of comparable questions on political issues, and on knowledge of the protest events. The availability of information on respondents awareness of the regional protest events allows us to assess how regional protests affected individuals awareness of, and support for, the demands of the protest movement among that sub-sample of respondents who admit to being aware of the national protest movement. By limiting our sample of respondents to individuals 2 For a detailed description of the survey questions and summary statistics see SI, Tables

14 who are aware of the protest movement we address concerns about ecological fallacy that are present in similar works on the cognitive effects of protests elsewhere (Branton et al. 2015; Wallace, Zepeda-Millán, and Jones-Correa 2014) and thereby increase confidence in our results. Drawing inferences on the effect of protest events on bystanders would be more challenging in the absence of items that specifically ask respondents whether they are aware of the protest movement or not. Our Protest-Event Dataset, assembled from the liberal namarsh.ru website, covers 428 anti-regime political protests events that took place across Russia during the electoral protest wave between 4 December 2011 and 31 May For each protest event, the dataset identifies the date, location and number of participants. Electoral protests during this period occurred across the nation. Even regions hitherto known for delivering a pro-kremlin result in elections and traditionally featuring low levels of political openness, comparatively smaller pools of activists and lower levels of pro-democratic attitudes experienced street activism. 3 Although territorially dispersed, the events featured a clear target for blame: they accused the political regime of abuses and called for political change in the country. In the early winter and spring months of 2012, the slogan Russia without Putin was echoed across Russia s eleven time zones. 3 As we show in the SI (Section 4.1) the correlation between regional democracy scores from the Democracy Index devised by prominent political geographers Nikolay Petrov and Alexei Titkov, and the frequency of electoral protests taking place during this period, is relatively weak. Press freedom, as well as other indicators of regional democracy in the country, which could be interpreted as proxies for latent dissatisfaction with the national government and the abuse of freedoms and rights traditionally fail to predict the occurrence and size of antiregime mobilization. 14

15 Figure 1, which presents how protest events were distributed across Russia, shades administrative regions witnessing a higher frequency of protest events in darker color. Figure 2 presents how the number of protests and protest participants fluctuated over time. In the time frame analyzed, the number of protesters varied from one to 100,000 (Moscow, 24 December 2011; and 6 May 2012). Despite this variance, nearly fifty protest events involved more than 2,000 participants. The number of protest events spiked around three key moments. In the immediate aftermath of the 2011 State Duma elections, forty-three protest events took place across Russia on December 10. On 4 and 19 February 2012, over 50 protest events took place across the country. Finally, protests took place in early March, as a response to President Putin s re-election. Figure 1. Variation in the Frequency of Protest Events: December 2011-May 2012 Source: Author Protest-Event Dataset. 15

16 Figure 2. Number and Size of Protests, December 2011-May 2012 Source: Author Protest-Event Dataset. In constructing our key independent variable, we follow Wallace et al. (2014) and calculate the number and size of anti-regime protest events taking place in each respondent s region between 4 December 2011 and up to the day of a respondent s interview. To alleviate the concern of systematic misreporting we use the mean number of protesters across two or three sources. Specifically, beyond namarsh.ru, we consult other regional websites reporting on protest events, such as the website of the left-leaning Institute of Collective Action (IKD), kasparov.ru and other online news sources. (For a similar approach, see Robertson 2013). We aimed to find at least one article that substantiates the details of each protest observation from namarsh.ru. Events not initially reported by namarsh.ru are added to the dataset only if they are referenced in an additional source. We also look for police estimates of the number of protesters reported on Russia s government websites. We expect regional protest events to be positively associated with awareness of, and support for, the national protest movement. In the second specification testing this hypothesis, we interact the protest variable with a proxy of political news consumption. The 16

17 interaction term helps us test the hypothesis that the cognitive effects of protests will be dampened among individuals who are more often exposed to the state s narrative on protests. In our third and final specification, we examine whether the use of repression against protesters systematically colors the effects of protests on public opinion. The suppression indicator we employ comes from our Protest-Event Dataset and captures active attempts to disrupt and disperse a protest and to arrest activists. We rely on this variable to construct a measure of regional suppression up to the day of a respondent s interview. All models presented below include the key protest variables and a battery of standard demographic controls. We also introduce a variable related to respondents participation in prior protest events, and control for voting behavior in the 2011 Duma Election. Survey items record whether individuals voted for the opposition, for the pro-kremlin United Russia party or abstained. Employment status is also controlled for, as are evaluations of household economic conditions, used here as a proxy for economic circumstances. A dummy variable that captures whether respondents reside in an urban or rural center is introduced to account for variance in political attitudes attributable to the effects of urbanization. The inclusion of regional fixed effects in the models allows us to ascertain whether protest effects should be attributed to variation in the frequency and size of protest events within regions. 4 Results and Discussion 4.1 Connecting regional protest events and awareness of the national protest movement As the first step in validating the claim that citizens information about the wave of electoral protests following the December 2011 parliamentary elections is largely influenced by regional protest events we perform analysis employing a measure of individual knowledge 17

18 about the events. We turn to one item in the Stephen White survey asking how much respondents know about the meetings and protests that took place in Moscow and other Russian cities after the elections to the State Duma. Based on individuals responses, we create a dichotomous Don t Know variable, as a measure of low knowledge about the protests (for a similar approach, see Gingrich 2014). Respondents who know nothing or almost nothing about the protests receive a score of 1, while the remaining respondents are assigned a score of 0. Evidence that regional protest activity enhances the overall visibility of the national protest movement should provide a solid foundation for the remaining analysis. The left-hand panel of Figure 3 shows a strong negative-bivariate correlation between the frequency of regional protest taking place up to the respondent s day of interview and the average of Don t Know responses (r= -0.40, p< 0.01). The right-hand panel of Figure 3 also points to a negative, yet weaker relationship between the regional per capita number of protesters taking to the streets up to the day of the respondent s interview and the average of Don t Know responses (r= -0.25, p< 0.01). To exclude the influence of the largest protests taking place in Moscow and St Petersburg, this graph limits the per capita number of protesters to just below 6,000 per region. The negative relationship between protester pool and ignorance about the national protest movement however holds, regardless of how we choose to code the size of regional protests. The bivariate correlation between the average in Don t Know responses and the size of all regional protests taking place across Russia in December and January is stronger and also negative (r= -.45, p< 0.01). 18

19 Figure 3. Bivariate Relationship between a Don t Know Response, Regional Protest Event Frequency (r= -.40, p <.01) and Size (r= -.25, p <.01) Note: The graph plots the bivariate correlation between the number (left-hand panel) and size (right-hand panel) of regional protests taking place up to the day of respondents interview and the average percentage of Don t Know responses. To calculate the size of protest events we divide the total number of protesters taking to the streets up to a respondent s day of interview with the total number of residents in the region (in thousands). We also run a logistic regression, controlling for the individual-level variables discussed in the data section of our paper, regional and survey fixed effects. Our statistical analysis is limited to individuals who reside in regions that experience unrest alone. Once more, the results suggest that the frequency and size of regional protest events is a significant 19

20 predictor of a Don t Know response (SI Table 3.1). When the frequency of regional protest events a respondent is exposed to increases from one to five, which is the mean number of regional protest events taking place in respondents regions between December and January, the probability of a Don t Know response drops from 0.28 [0.24, 0.32] to 0.25 [0.22, 0.27]. In late January, the probability that a Moscow resident experiencing a maximum of twenty protests will report a Don t Know response is by a magnitude of.12 [-0.20, -0.05] lower than that of a resident in proximity to a single event (SI Table 3.2). Put together, these findings lend strong support for the argument that the spread of anti-regime protests outside of the capital increases the visibility of the national protest movement. A potential objection to the conclusions that we derive from the statistical analysis, however, is that a region s culture of activism is likely to be correlated with the number and size of regional protest events and with regional public opinion. In politically liberal regions, one might argue, citizens are more likely to show awareness of the national protest movement and would be predisposed to support the demands made at local anti-regime rallies. Spatial variations in civic activism and democratic practices are pronounced in democracies (Almond and Verba 1963; Putnam 1993) and under electoral authoritarianism (Frye, Reuter, and Szakonyi 2014; Gel man and Lankina 2008; Lankina and Getachew 2012; Lankina, Libman, and Obydenkova 2016a, 2016b; Lankina and Skovoroda 2016; Moraski and Reisinger 2014). To address these concerns early on in our work, we instrument the frequency and size of protest events by exploiting variation in weather conditions across regional capitals. Concretely, we use temperature on the day of regional rallies as an exogenous source of variation in protest activity. Exploiting variation in weather conditions is a well-established instrument in studies that explore the effects of protest on political change (Collins and Margo 2004, 2007; Madestam et al. 2013). Drawing on these works we assume that weather 20

21 conditions affect attitudes toward the protest movement only through the frequency of protests and size of protesters. The instrumental variable captures the day temperature in the regional capital for every regional protest event that takes place between December 2011 and January We obtain this information from Russia s meteorological service. In the models presented below, we use the average temperature across regional rallies that take place up to the day of a respondent s interview. Our expectation here is straightforward: colder weather could reduce protest attendance on the one hand and the occurrence of spontaneous protest gatherings on the other. The first-stage results from the instrumental variable regressions presented in Table 1 below confirm this to be the case (Models 1.1 and 2.1). Figure 3.1 in the SI also points to a positive correlation between daily temperature and protest attendance across events taking place between December 2011 and January The second-stage IV regressions presented in Table 1 below (Models 1.2 and 2.2) show that regional protest events are associated with the visibility of national protests. The higher the frequency of regional protest activity, even in the early two months of the protests, the less likely respondents are to report that they know nothing or almost nothing about the wider national protest wave (Model 1.2). And, although the protest size coefficient in Model 2.2 fails to reach statistical levels of significance it remains in the direction anticipated. 21

22 Table 1: Instrumental Variable Regressions (1.1) (1.2) (2.1) (2.2) First Stage Second Stage First Stage Second Stage Protest Frequency Protesters Per Capita Regional Temperature 0.728*** 1.499*** (0.0728) (0.109) Protest Measure * (0.0414) (0.0282) Urban Settlement 1.470*** 0.283** 1.618*** 0.232** (0.198) (0.118) (0.210) (0.118) Federal Regions (Districts) FE Yes Yes Yes Yes Demographic Controls Yes Yes Yes Yes Constant 10.76*** *** ** (0.830) (0.500) (0.856) (0.359) Wald Test Observations 1,160 1,105 Notes: The table reports IV-probit coefficients obtained from the first and second stages of two instrumental variables models. Standard errors are given in parentheses *** p<0.01, ** p<0.05, * p<0.1. The models control for demographic characteristics: prior protest behavior, education, gender, age, ethnicity, media exposure, voting behavior, household conditions and employment status. Full results are presented in Table 3.3 in the SI. These results are in line with earlier work by Madestam et al. (2013) in that they show that regional anti-regime protests do not simply reveal pre-existing political preferences but are in a position to bring about changes in public opinion. Relying on the protest frequency measures and confident about the link between regional protest and public opinion, we now move on to examine how protests affected awareness of electoral fraud and support for the protesters between December 2011 and May

23 4.2 Regional protests, awareness of fraud and support for the demands of protesters The models presented in Table 1 predict awareness of fraud and support for the demands of protesters as a function of regional protest events taking place up to the day of a respondent s interview. We now analyze whether in regions that witnessed rallies, attitudes towards the protesters and their demands will vary as a function of the number of these events. We estimate Model 1 using OLS analysis because the dependent variable measures perceptions of electoral falsifications on a five-point scale, with higher values denoting that the respondents think that there had been great unfairness in the 2011 State Duma elections. We estimate Model 2 using ordered logistic regression due to the ordinal nature of the dependent variable. Our dependent variable, which captures attitudes towards the protest movement, is measured on a scale that ranges from 1 to 4. Higher scores of this variable denote that respondents are strongly supportive of the protesters demands. For both questions, we restrict the sample of respondents to those who indicated awareness of the protest events that had occurred in their region prior to the survey. The results presented in Model 1 indicate that holding everything else equal, for every additional regional protest taking place up to the date of a respondent s interview, the likelihood that he or she will report higher levels of electoral falsifications in the State Duma election increases by a magnitude of ~ This effect is small, however, during this period, considering that citizens in regions that did experience unrest were exposed to an average of nine protest events up to the day of their interview. At nine protest events, which is the regional average, the effect of protest increases to ~.07. Similarly, for respondents in Moscow, where by May over eighty protest events occurred, the effect of protest exposure increases to ~.71. This effect is substantively important, considering that the variable capturing perceptions of electoral unfairness is measured on a five-point scale. The protest 23

24 variable, however, fails to reach statistical levels of significance in Model 2. The results at this stage therefore suggest that by the end of the protest wave, regional protest events had been unable to generate or indeed to maintain the support of regional publics. The statistically significant coefficient from the survey fixed effects, however, suggests that support for the demands of the protesters, as well as perceptions of electoral falsifications, fluctuated significantly over the protest wave. In line with survey evidence from Russia s widely respected public opinion survey agency, the Levada Center, the results presented here suggest that by spring, protesters enjoyed lower support among the public than they did in the early winter months (Levada 2012). When we repeat the analysis using the January survey alone, the regional protest events coefficient is positive and statistically significant at the 1 percent level. Holding all other covariates at their means, every additional protest event that took place over December 2011 and January 2012 increased the likelihood of respondents strongly agreeing with the demands of the protesters by a magnitude of.008 (SI Table 3.4). Turning to the controls, we find that only a handful of significant variables emerged. Prior participation in protest events of any type over the course of the last two years predicts higher perceptions of electoral injustice and support for the protesters. Individuals residing in urban, in contrast to those residing in rural, settlements report higher levels of electoral unfairness, an effect which is statistically significant at the 5 percent level, but are no more likely to support the demands advanced by protesters. Exposure to state-controlled media does not have a statistically significant effect on either perceptions of fraud or on attitudes towards the protesters. As one would theoretically anticipate, when compared to opposition voters, supporters of the pro-kremlin United Russia party are less likely to report higher levels of electoral fraud and to share the protesters demands. Voters who abstained from voting in the 2011 election are likewise less likely to support the demands of the opposition in Model 2. Individuals who report no improvement in their household conditions are equally 24

25 more likely to perceive the 2011 State Duma elections as fraudulent. They are also more likely to support the broad demands advanced by the protest movement. 25

26 Table 1. The effect of regional protests on perceptions of fraud and political attitudes (1) (2) Electoral Unfairness (Linear Regression) Supports Demands (Ordered Logit) Regional Protest Frequency 0.009*** (0.003) (0.004) Protested Previous 2 years 0.331** 0.743** (0.154) (0.319) Secondary Education A (0.071) (0.096) Higher Education (0.069) (0.131) Male (0.065) (0.105) Age (0.002) (0.003) Russian (0.077) (0.186) Urban Settlement 0.229** (0.089) (0.235) Political News Watching (0.096) (0.169) Voted UR B *** *** (0.105) (0.126) Abstained *** (0.074) (0.144) Household: Same C 0.225*** 0.411*** (0.072) (0.154) Household: Deteriorated 0.544*** 0.520** (0.133) (0.235) Out of Labor Force D (0.078) (0.134) Unemployed (0.110) (0.286) Survey Fixed Effects *** *** (0.146) (0.282) Region Fixed Effects Yes Yes Constant 2.784*** *** (0.269) (0.471) Observations 2,123 1,660 R-squared Notes: Robust standard errors, clustered by regions in parentheses *** p<0.01, ** p<0.05, * p<0.1. The baselines are: A: Primary Education B: Voted for the Opposition; C: Household Evaluations: Improved: D: Employment Status: Employed. 26

27 4.3 Regional Protests and Regime Strategies Media Effects Our theoretical framework anticipates that the effect of regional protest on public opinion will be contingent on citizen exposure to state-controlled media and upon the use of police repression against protesters. We therefore begin by exploring whether the cognitive effects of protests are conditional on the coverage of the protest movement in national media. We already know that between December 2011 and May 2012, electoral protests represented a salient and highly visible issue in state-run media. Coverage of the protest movement, however, primarily sought to prevent the erosion of popular support for the regime. Statecontrolled news sources underestimated the size of anti-governmental protests; muted the demands that Russia s top leadership should be held accountable for electoral misconduct (Koesel and Bunce 2012; Smyth and Oates 2015); and employed frames of violence, disorder and unrest in portrayals of the protest movement (Lankina, Watanabe, and Netesova 2016). The media variable that we employ for this part of the analysis, which captures whether and how often individuals watch political news on television, is meant to proxy for the effect of media coverage of protests on political preferences and attitudes. This variable is coded as 1 if respondents frequently watch political news on TV and as 0 if respondents seldom, or hardly ever do so. To test whether news media consumption affected the cognitive effects of protests, we interact this dummy variable for media exposure with our main protest indicator. Figure 4 below aims to ease the interpretation of the interaction terms. The upper panel of this figure presents adjusted probabilities from a model predicting perceptions of electoral falsifications in prior elections, while the lower panel predicts support for the demands of the opposition. Full results are presented in Table 3.5 in the SI. 27

28 The predicted probabilities in the top panel of Figure 4 indicate that for both groups of respondents, that is, those who watch political news and those who don t, perceptions of electoral unfairness vary as a function of the number of prior regional protest events. Among the group of respondents who often watch political news, the probability of reporting higher levels of electoral unfairness increases by.07 as the number of prior regional protests increases by one to nine, that is, the average number of regional protest events taking place ahead of the respondents day of interview (Table in the SI). Perceptions of electoral unfairness, however, are consistently higher among respondents who hardly ever or never watch political news. At a single protest event alone, for example, the probability of reporting high levels of fraud, or unfairness is by a magnitude of.06 (-.28,.16) higher among those respondents who are not exposed to the coverage of the protest movement in the national media, although the comparison between the two groups fails to reach statistical levels of significance. The predicted probabilities presented in the lower panel of Figure 4 illustrate that exposure to political news dampens protesters ability to attract the support of the wider citizenry. Respondents exposed to the state s rhetoric when they watch TV are less responsive to fluctuations in regional unrest as their attitudes do not vary as a function of the number of regional protests. Conversely, the probability that individuals who hardly ever or never watch political news will come to agree with the demands of the protesters increases by ~.02 (.00,.04) as the number of regional protest events increases from one to nine. 28

29 Figure 4: TV news consumption, protest events and public opinion Notes: The figure is based on the results presented in Table 3.5 in the SI. The figure plots the difference in predicted probability of reporting higher levels of electoral falsifications (upper panel) and agreeing with the demands of the protesters (lower panel) when the frequency of regional protest events increases from one to nine for two groups: those who frequently watch political news on TV and those who rarely do so. The figure shows that the consumption of political news reduces protest events ability to generate support among the wider regional citizenry but does not strongly condition perceptions of electoral falsifications. In producing the predictions, we hold all other covariates at their means. The vertical bars represent 95 percent confidence intervals. 29

30 One potential concern about these results is that individuals with pre-formed opinions about electoral fairness and the demands of the protesters could self-select in terms of how often they will be watching political news across a range of TV outlets. Accordingly, one may anticipate that citizens who watch political news on TV more often would be also more likely to disagree with the demands of the opposition and to report lower levels of fraud. We therefore perform several additional tests to probe the robustness of our media exposure results. In the analysis presented in the Appendix (SI 4.2) we show that respondents who frequently watch political news on TV, and those who seldom do so, are barely distinguishable in terms of their political preferences and prior participation in protests. Similarly, using multivariate analysis we show that individuals who watch political news on TV (as opposed to those who don t), or those who watch news more frequently (as opposed to those who watch news less frequently), do not hold more authoritarian views, even when three different specifications of the media exposure variable are employed. Finally, it is worth noting that in the Stephen White Survey, 93 percent of those respondents who claimed to frequently watch political news reported watching the news on Russia s First Channel. Although this percentage was slightly lower among respondents in the RES (85 percent), these results suggest that the overwhelming majority of respondents who frequently watched political news obtained them from state-controlled media. In sum, respondents more frequently exposed to the state s rhetoric via the mass media are less responsive to fluctuations in regional unrest as their support for the protest movement does not vary as a function of the number of protests. Therefore, and in line with existing scholarship (Chong and Druckman 2007a, 2007b; Druckman and Chong 2013), our results suggests that when the regime and the protest movement offer alternative interpretations of the same issue, the state s narrative on protests has the potential to dampen demonstrators ability to win the hearts and minds of the wider citizenry. 30