PÁL SUSÁNSZKY Demobilization Processes: Perceived Risks of Protest Participation among Hungarian Students

There is abundant literature on individual-level characteristics that encourage citizens to participate in political demonstrations. However, empirical studies on demobilization and factors that prevent people from joining protests remain scarce. In this paper, I zero in on the perceived risks of political participation. Two questions are examined: first, how protest willingness is shaped by perceived risks, and second, what political and socio-economic factors explain risk perception. I answer these questions using the representative sample of 800 Hungarian university students from the Active Youth Survey (2019). Hungary has a special position in Europe because it is defined neither as a liberal democracy nor as sheer autocracy, but an ‘illiberal regime’. In non-democratic illiberal societies the state does not apply overt repressive techniques against dissident groups, although protest participation is still not a riskless form of political action, as regarded in developed democracies. I apply logistic regression models to predict both protest willingness and perceived risks of protest. Results confirm the importance of risks in extra-parliamentary protest politics, since almost half of the university students see their participation in demonstrations as somewhat risky. Regression models show that perceived risks are to some extent politicized, but risks have their own significant role in explaining protest (un)willingness.


Introduction
Recent review articles (Earl, 2011;Honari, 2018) about state repression, and the threats and risks of political participation emphasize that there is hardly any study on the link between political participation and its risks. Individual-level analyses are rare, although in the last few years the topic has attracted some scholarly attention (Young, 2019;Curtice and Arnon, 2019;Ayanian and Tausch, 2016;Ayanian et al., 2020).
Theories of collective action are based on the idea that there are factors that make political participation easier, and others that impede them, or at least make political action more difficult. The first group of factors mobilize, while the others demobilize citizens.
Research results however show a mixed picture on the risk-protest nexus. Qualitative studies demonstrate that social movement activists assess various forms of risks (e.g. Boykoff, 2007;Della Porta, Fillieule and Reiter, 1998;Earl, 2011) that may hinder their protest participation. On the other hand, there is surveybased evidence of a positive correlation between perceived risks and participation (e.g. Opp, 1994;Ayanian and Tausch, 2016). Scholars have shown that the riskier it is considered to be active in demonstrations, the more willing respondents are to participate. Opp and Roehl (1990) offer us a concise explanation of the puzzle: perceived risks not only deter people but also invoke dissatisfaction, anger, or other political attitudes that increase protest willingness.
In this paper, I study how perceived risks predict willingness to participate in protest participation, and I also analyse how socio-economic and political factors shape risk perception. To answer these research questions, I use the representative sample of Hungarian university students (the fourth round of the Active Youth Survey (2019), N=800) and apply logistic regression models to predict both protest willingness and three different types of perceived risks, namely friends' disapproval, counter protesters' and police attacks.
Some of the existing research on the link between perceived risks and protest participation was conducted in non-democratic countries like Egypt (Ayanian and Tausch, 2016), Zimbabwe (Young, 2019), or the communist East Germany (Opp and Roehl, 1990), where engagement in protests was regarded as risky, and where physical retaliation or state repression were part of everyday politics. Other studies focus on the repressive techniques applied by democratic states against dissident groups (e.g. Almeida, 2018;Earl, 2011), and focus mostly on activists on the political fringes (e.g. Boykoff, 2007;Linden and Klandermans, 2006).
Hungary has a special position in Europe because of its obvious backlash in democracy (Buzogány, 2017;Bogaards, 2018) and the increasing state repression against civil society organizations and academia (Enyedi, 2018;Gerő et al., 2020;Kuti, 2016). As an illiberal state, it aims to control both parliamentary and street politics (e.g. Robertson, 2010;Cheng, 2016). However, Hungary is not an autocracy, where state represses government-critical demonstrations. Incarcerating and INTERSECTIONS. EAST EUROPEAN JOURNAL OF SOCIETY AND POLITICS, 6(4): 63-92.
shooting demonstrators is not happening in Hungary, but protest participation is still not risk-free.
In this article, I do not analyse the state's demobilizing techniques, but will show the attitudes university students have towards extra-parliamentary politics. I will also look into how risky protesting is in the eyes of the new generations of the intelligentsia, who have socialized in non-democratic Hungary.
Unsurprisingly, Hungarian university students are more active and more willing to participate in demonstrations, as compared to the whole population. One-fifth of students see the risk of peers' negative reactions, and half of them see police attacks as a kind of risk that would likely follow their participation. Regression models reveal that perceived risks predict protest willingness in a complex way. Physical and non-physical risks could predict protest willingness both negatively and positively, which indicates that, as Opp and Roehl (1990) suggest, there are different mechanisms linking risks to willingness. An alternative explanation may be that risks are only proxies of political identities, party preferences and political orientations, thus risk perception is politicized and fully shaped by these political factors. Results of the analyses, however, clearly refute this reasoning.
The study is structured in the following manner. First, I delineate how micro-theories focusing on the individual level explain why people participate in protest activities. Second, I discuss theories and empirical studies about the costs and risks of protest participation. In the third section, I describe Hungary's illiberal political context. Fourth, I introduce my hypotheses derived from the literature. Finally, after presenting the results, I discuss the findings and show that risks could demobilize but also encourage university students to take part in demonstrations against the government.

Micro theories of protest participation
Research on political participation has a long tradition in the social sciences. How can we explain that some citizens have their voices heard, while others do not? Why do people vote, take part in protests, sign petitions or, in more general terms, join collective actions?
Explaining political participation lies at the crossroads of many disciplines. Sociology, economics, social psychology, as well as social movement studies have made efforts to elucidate the above questions. However, all of these branches of social sciences have developed their own theoretical models. Although these disciplines differ in their approaches, and their models focus upon different aspects of this problematic, they all conclude that while some social and psychological factors mobilize citizens, others demobilize them.
Mobilizing factors, on the one hand, are incentives and benefits (e.g. Mueller and Opp, 1986), grievances (e.g. Pinard, 2011), resources (e.g. Verba, Schlozman and Brady, 1995), protest efficacy (e.g. Saab et al., 2015, Opp andKittel, 2010) and political values (Dalton, van Sickle and Dalton, 2010) that push citizens towards participation. On the other hand, demobilizing factors are forces that hold people back from political activism. These are the individually estimated costs and risks of demonstrations. The general -and simplified -model of an individual decision about political action is: A=B-C. In other words, the probability of participation (A) hinges on the mobilizing (benefits of the action) and the demobilizing (costs of the action) factors 1 .
While mobilizing factors have always been in the limelight of political sociology and political psychology, demobilizing factors are still under-researched (Earl, 2011;Honari, 2018). I see three main reasons why scholars have been reluctant to research citizens' risk perceptions over the last three decades: Firstly, our theories aim to explain participation rather than nonparticipation or passivity. Micro-mobilization theories differentiate between protest participants and non-participants, and try to dissect 'personal characteristics [that] determine which individuals are most likely to protest within a nation' (Dalton, van Sickle and Weldon, 2010: 56). If we can explain why citizens participate in demonstrations, we can also explain why others do not. Those who are dissatisfied with the government are more inclined to protest, which indicates that the less dissatisfied are less willing to protest. Thus, our explanation shows a symmetry between participation and non-participation. As Ward (2016) states, '[i]dentifying attributes participants possess and non-participants lack is the sine qua non of this sociological literature' (Ward, 2016: 854). However, others regard demobilizing factors as a separate and substantive element. According to Goldstone and Tilly (2001), threat is an independent factor that influence both dissident mobilizations and also state reactions. In their analysis threat is defined as a demobilizing factor instead of less opportunity to protest, or lack of motivating attitudes.
Secondly, there is also a technical reason behind the moderate intensity of empirical research, namely, the lack of a standardized and tested questionnaire battery for perceived risks of protests. While, for example, internal and external political efficacy was introduced in the early 1950s (Campbell, Gurin and Miller, 1954), and other relevant incentives for political participation (e.g. dissatisfaction with the government and left-right political attitudes) are quantified by standardized and refined survey questions, negative incentives, costs, and risks have less developed measurement techniques.
Finally, in mature democracies state repression, threat, and political demobilization seem to be less prevalent than in non-democracies. Thus, sociologists in the western part of Europe have not perceived political risks as a relevant social phenomenon. If something is not present in a society, and 1 See the much more sophisticated models of voting by Riker and Ordershook (1968), and Blais (2000), or of protest participation by Muller and Opp (1986), Klandermans (1984), and Goldstone and Tilly (2001). Some of these studies are based on the collective action theories and the seminal book by Mancur Olson (2009Olson ( [1965).
These models are extended with subjective evaluations of probabilities: for instance, the probability that one's participation in the demonstration will be decisive for the success of the collective action; or the evaluation of others' participation.
examining it is not necessary either, then researching it is not very likely to happen.

Perceived risks and costs of protest participation
As noted above, research on protest demobilization is scarce, but is not completely missing from the literature. McAdam (1986) emphasizes that at the level of individuals, it is important to distinguish between the costs and risks of participation. The cost of an action is defined by the time, money, and energy devoted to participation. Risk, on the other hand, 'refers to the activists' subjective anticipation or expectation of a cost that they may incur as a result of their movement participation' (Wiltfang and McAdam, 1991: 989). The literature on protests identifies two main types of risks citizens may face (e.g. Davenport, 2005): (1) physical retribution by police, other state actors, and counter-demonstrators, and (2) non-physical risks. Physical risks include being arrested, beaten, injured, tortured, or killed (Wiltfang and McAdam, 1991;Opp, 1994), abducted, having one's property destroyed (Young, 2019), and being sexually harassed or abused (Ayanian and Tausch 2016;Young 2019). On the other hand, protesting may be associated with negative non-physical consequences, such as losing one's job (Shriver, 2000), 'problems for close family members' (Opp, 1994), or losing one's social contacts (e.g. severing ties with friends or relatives) (Tóth and Kertész, 2016).
According to collective action theory (e.g. Olson, 2009;Opp, 2009), costs and risks prevent participation in demonstrations through decreasing people's willingness to protest, and therefore costs and risks are deemed as demobilizing factors. Since previous, mostly qualitative, research has revealed various types of risks, my quantitative survey needs to cover some of these different dimensions. In the questionnaire we asked our respondents about four kinds of risks: (1) disapproval of friends and relatives, (2) trouble at work or at university, which count as non-physical risks, and (3) attacks by counter-protesters; (4) police attacks, which are physical risks. I chose these four types of risk, because they are diverse in their consequences, and also because as I will show in the next sections, since 2010 the Hungarian media has been intensively reporting on these risks.
In order to make the classification of demobilizing factors clearer, Figure 1 summarizes the categorization. Demobilizing factors can be broken down to costs and risks, and within risks we can distinguish between physical and non-physical risks. In our survey, two items represent physical, and two others non-physical types of risks.

Previous research on perceived risks and protest participation
There is some empirical evidence that protest participation has negative personal consequences even in well-developed democratic contexts. Qualitative research based on interviews with activists and on historical analyses demonstrates that social movement activists even in Western democracies often face a wide range of physical (e.g. Boykoff 2007;Della Porta, Fillieule and Reiter, 1998) and nonphysical risks (e.g. Boykoff, 2007;Marx-Ferree, 2004;Shriver, 2000) that may hinder their protest participation. These studies, however, mostly focus on extremist groups (e.g. Linden and Klandermans, 2006) and violent dissident groups (e.g. Boykoff, 2007;Della Porta, Fillieule and Reiter, 1998), rather than on nonviolent demonstrations of ordinary people. Researchers also have found a negative association between risks and activism in countries where state repression and the possibility of severe injuries in protest events is overt and obvious for every INTERSECTIONS. EAST EUROPEAN JOURNAL OF SOCIETY AND POLITICS, 6(4): 63-92.
citizen. For example, in China (Deng and O'Brian, 2013), in South Korea in the early 1970s (Chang, 2015) or in Zimbabwe (Young, 2019) to mention just a few.
Surprisingly, there is survey-based evidence for a positive correlation between perceived risks and participation (e.g. Opp, 1994;Ayanian and Tausch, 2016). Analysing survey data from West and East Germany in the 1980s, Opp and his colleagues (Opp and Roehl, 1990;Opp and Gern, 1993;Opp, 1994) found that the riskier it is considered to be active in protest, the more willing respondents are to participate in demonstrations. Ayanian and Tausch found the same positive correlations in the case of protesters in Egypt (Ayanian and Tausch, 2016) and also in Russia, Hong-Kong or Turkey (Ayanian et al., 2020).
How can we explain these seemingly contradictory results? Opp and Roehl (1990) suggest that political repression shapes political activism through two parallel mechanisms. The first is the deterrence mechanism, a direct negative effect of repression on participation. On the other hand, there is the radicalization mechanism, an indirect path through which repression triggers activism. Not only does the indirect effect of repression increase the perceived levels of risk, but it also produces attitudes (e.g. moral incentives, anger, and group efficacy) that make supporters more likely to participate.
In this study, I examine the risks of protest as perceived by Hungarian university students, because Hungary as an illiberal member state of the European Union is neither a democracy like Western European countries, nor an autocracy like Zimbabwe. Incarcerating and shooting demonstrators is not happening in Hungary, yet as we found, in 2014 citizens (and not the zealous extremists) saw demonstrations as somewhat risky (Majtényi, Kopper and Susánszky, 2019).
Theories on protest participation aim at explaining political protest mobilizations in the most general form, and providing understanding of political processes across different political contexts. Mobilization in high-risk political contexts is much more difficult than in well-developed democracies, where lowrisk activism is the most dominant form of political protests. However, participation can be dangerous and risky in democratic countries as well. 2 Thus it seems necessary to put further effort into dissecting the risk-protest nexus, and extend our knowledge on demobilizing processes in non-authoritarian regimes.
As the next section explains, there is a vivid public discourse in the media on state repression and risks of political participation.

Risks of protest participation in illiberal Hungary
Prime Minister Viktor Orbán first used the term 'illiberal state' for Hungary in a 2014 public speech. 3 Some government leaning journalists and analysts interpreted the term as 'post-liberalism,' or 'national democracy.' 4 However, most social scientists have been reading Orbán's speech and his politics as a democratic backlash. Analyses cover the centralization of the media (Polyák, 2019), changes in the electoral law (Papp and Zorgit, 2018), rewriting the constitution (Batory, 2016; Várnagy and Ilonszki, 2017), and the weakening role of opposition parties in parliament (Várnagy and Ilonszki, 2018). All the studies show that the governing Fidesz party has been extremely successful in power centralization, which makes it easy for them to control institutionalized politics.
As part of the Orbán regime's centralizing politics, the civil sphere has been restrained, since civil society organizations, and especially those who work for human rights protection, perceive shrinking political opportunities (Gerő et al., 2020). Moreover, a significant part of Hungarian civil society actors, particularly those who have criticized the government, are intimidated and stigmatized (Freedom House Country Report, 2018; Maerz et al., 2020). Beyond demobilization at the level of organizations, it is worth analysing how ordinary citizens assess risks of protest participation.
In the present political situation, when opportunities for party politics are narrowing, political protests are gaining a special role. This is not so because political decisions could be more effectively influenced through demonstrations, but because political protests have the function to sway public opinion, social value systems, and to build the opposition's group identity (Amenta and Young, 1999). However, we know much less about the way illiberal states manage to control such non-institutionalized forms of participation as protesting and organizing demonstrations. As Robertson correctly argues, non-democratic regimes do not want to eliminate competition; on the contrary, it is rather 'something that they consciously allow and try to control' (2011: 13). Nondemocratic governments show off public demonstrations as a testimony of freedom of speech and expression, and a limited scale of protests informs the regime about grievances within the society, but they want to be able to react before discontent escalates (Lorentzen, 2013).
To illustrate how public perceptions of risks have evolved in the last few years, I have collected news content from Hungarian mainstream news portals. 5 The main goal of the following non-systematic analysis of Hungarian media outlets is to show that public discourse covers the issue of risks of protest. Moreover, as Gamson (1992) points out, in addition to general public opinion and personal experiences, media content is crucial in shaping citizens' political opinions. Thus, we can assume that Hungarian university students, as part of Hungarian society, are also exposed to such influences.
In the media I have found four main types of risks regarding protest participation: (1) non-physical risks of repercussions at work, (2) friends' disapproval, (3) physical threats of counter-protesters and (4) police attacks.
Some Hungarian media report conflicts at work. Employers do not always like their employees' political activism and their open confrontation with political power. For example, the director of a state founded think-tank sent warnings to his employees not to like or post any Facebook content against the government's 'Olympic Budapest 2024' campaign. 6 The woman who made this letter public was subsequently dismissed. 7 In another case, a cook was fired from the high school where he worked. His dismissal happened shortly after he had attended and held a speech in a protest against the government's measures and communication against civil organizations. 8 A few weeks later, after high school students had protested in Budapest against the unjust educational system, a 17-year-old protester's home was searched and his computer was confiscated. 9 Friends' negative reactions is another non-physical risk citizens are faced with. I did not find reports of that type of risk in the media, however researchers have documented cases of peers' negative reactions. A few years ago, a group of sociologists did interviews with volunteers and activists who in 2015 aided refugees near the Hungarian border. Tóth and Kertész cite an activist saying that helpers in the refugee crisis 'all had confrontations with their environment, family, acquaintances, colleagues' (2016: 116). Another 22-year-old woman said that Facebook acquaintances had broken ties with her due to her political activism (Tóth and Kertész, 2016: 302). These stories tell about the non-physical risks of protest participation or political activism at work.
As I explained in the theoretical sections, the effect of perceived risks is not obvious. The deterrence mechanism could demobilize potential participants, but through the indirect effect of the radicalization mechanism they may be spurred to protest.

Hypotheses
According to collective action theories (e.g. Opp, 2009), perceived risks decrease protest willingness. However, observational surveys have found that, contrary to expectations, perceived risks (through the radicalization mechanism) might increase the inclination to participate in protest (Ayanian and Tausch, 2016;Opp, 1994). Thus, I hypothesize that higher perceived risks are associated with higher willingness to protest. H1: Both physical and non-physical forms of perceived risks positively predict protest willingness.
There seems to be no research so far to examine how perceived risks of protest participation are formed by social background (e.g. father's education, and the family's subjective financial status), by political attitudes (e.g. leftist or rightist ideology), or by party preferences. However, there are several studies about risk perception regarding terrorist attacks (Huddy et al., 2002;Huddy et al., 2005;Skitka, Bauman and Mullen, 2004;Lerner et al., 2003), natural disasters and global warming (Armas, 2006;Mayer et al., 2017;Sun and Han, 2018;Sund, Svensson and Andersson, 2015;Vasquez et al., 2018) and nuclear energy (Opp, 1986;Sjöberg, 2004;Sjöberg and Drottz-Sjöberg, 2009). Interestingly, irrespective of the source of threat, the literature is almost consistent in assessing the role of demographic background in risk perception. Studies in various countries have found that women and less educated people report higher levels of risks (Armas, 2006;Huddy et al., 2002;Huddy et al., 2005;Mayer et al., 2017;Sjöberg, 2004;Skitka, Bauman and Mullen, 2004). In the case of environmental and nuclear threats, age is a significant predictor: older people perceive greater risks of earthquakes (Armas, 2006) and nuclear waste (Sjöberg, 2004).
Recent studies show that ideology plays a significant role in protest participation in Central and Eastern Europe (Borbáth and Gessler, 2020;Kostelka and Rovny, 2019). Perceived risks of protests may act as proxies for party preference and ideology. The risk assessment of those who believe that police attacks are more likely may reflect their party or ideological preferences, which may be driving protest propensity.
Due to the exploratory nature of my research, I formulate my hypotheses for the explanation of risk perception in a general manner. I expect that perceived risks are shaped by respondents' political attitudes, party preferences and socioeconomic background.
H2.1: Socio-economic status predicts perceived risks of protest participation negatively. Namely, respondents with lower social status (lower level of father's education, and lower subjective economic position) sense higher levels of risks.
H2.2: There are gender differences between male and female university students in perceived risks of protest participation. I hypothesize that women perceive greater risks than men.
H2.3: Perceived risks are shaped by political attitudes of left-right, liberalconservative, moderate-radical orientations, satisfaction with democracy and interest in politics.
H2.4: Perceived risks are shaped by party preferences. I hypothesize that Fidesz supporters assess protest participation as less risky, whereas supporters of oppositional parties assess it as riskier.

Data and methods
Students are considered a highly important social group for social movement studies since young people at university are prone to take part in demonstrations (McAdam, 1986;Sloam, 2013;van Dyke, 1998;Schussman and Soule, 2005).
Over the last decade, young Hungarians at high schools and universities have been very active in political demonstrations. They organized protests against the government's education policy and university fees in 2011 (Gerő and Susánszky, 2014a;2014b), and in 2016 against the former education secretary who called teachers 'dishevelled and unshaven types in checked shirts.' 14 For testing the hypotheses, I use the fourth round of the Active Youth Survey conducted among Hungarian university and college students in February 2019. 15 The early months of 2019 were characterised by relative political calm, since national elections had been held in 2018, and the European Parliamentary election campaign only started in March 2019 (Susánszky and Kritzinger, 2020). Our quota sample (N=800) is representative of gender, faculties (e.g., Medicine, Arts, Humanities, Social Sciences, and Natural Sciences) and level of education (BA, MA, and PhD). Interviewers applied the random walk selection method within the campus, following the strict prescriptions of randomization and selection. More than a hundred students 16 helped the fieldwork as interviewers who conducted face-to-face interviews with their fellow students. The interviews lasted 22 minutes on average. The topics covered by the questionnaire included social background, political socialization in the family and at school, plans of emigration, political attitudes, democratic values, political activity, party preferences, and one section of the questionnaire focused on protest participation. We measured willingness to protest participation with the following question: 'Would you do or would you not do any of the following to protest against a government action you strongly opposed?"' 17 Respondents answered on a six-point scale (1='I definitely would not' and 6='I definitely would'). After the willingness question, respondents were asked about perceived risks of protest participation: 'If you decided to participate in a demonstration against one of the government's actions you strongly opposed, in your opinion, to what extent would you risk that...' 1) your friends, relatives, and acquaintances might reprimand you due to your participation. 2) you might face repercussion at work or at school due to your participation. 14 https://www.smh.com.au/world/checked-shirts-begin-to-haunt-hungarian-authorities-20160325-gnr8qa.html (Accessed: 26-11-2019). 15 Principal investigator: Andrea Szabó; more information about the project: http://www.aktivfiatalok.hu/. 16 The interviewers are sociology, political science or social sciences majors, thus they have all taken quantitative methodology courses. In addition, they attended an interviewer training session, supervised by one of the three senior researchers.

3) you might be attacked by counter-protesters 4) you might be attacked by the police.
Respondents evaluated the risks listed above on a seven-point Likert-scale (1= 'not at all' to 7='very much'). 18 The four items cover the two main dimensions of risk perception: non-physical and physical. Friends' negative reactions, disapproval and negative consequences at university (and at work) are non-physical risks. Police and counter-protesters' attacks count as physical risks that may result in physical injuries. The reliability check shows (Cronbach's alpha=0.64) that the four items do not provide a consistent scale of perceived protest risk. Presumably, this is because the four items grasp multiple dimensions of risk. Therefore, I do not aggregate them as a single factor, but analyse the four items separately.
Since both the variables measuring perceived risks and the protest willingness items are non-normally distributed, 19 I recoded them into dummy variables. In the case of the four perceived risk items, I denoted a low level of perceived risks (0) if the risk scale value was lower than 5, and a high level of perceived risks (1) if the risk value was above 4. In the case of willingness to protest, however, values of 1 to 3 were recoded into 0, while values from 4 to 6 into 1. Thus, 1 denotes strong willingness, and 0 means weak or no willingness to participate in government-critical rallies.
For measuring socio-demographic characteristics and economic status, I used the following variables: • gender (male or female) • father's level of education (primary, secondary, or tertiary level) • subjective economic status (less than adequate, just adequate, or more than adequate) • place of residence (rural, city, or Budapest) • level of education (BA, MA, or doctoral studies) For measuring political attitudes, I used ideological orientations (left-right, liberalconservative, and moderate-radical) measured on seven-point scales, dissatisfaction with the working of democracy in Hungary (dummy variable 0 denotes 'satisfied', 1 denotes 'dissatisfied'), and political interest (dummy variable 0 denotes 'not interested', 1 denotes 'interested in politics'). The ideological  2016)) on an online sample. However, in the Active Youth Survey we have changed the wording and the range of the scale. 19 According to the Shapiro-Wilk normality test, the distribution of these variables is significantly different from normal distribution. Test statistics for protest willingness, perceived risk of friends' disapproval, repercussion in university classes, counter-protesters and police attacks is W=0.92, W=0.84, W=0.87, W=0.93, W=0.91, respectively, p values belonging to the statistics are lower than 0.001. orientation variables are standardized; thus, the scale has a mean of zero and a standard deviation of one.
To define party preferences, I used the following question: 'Which party would you vote for if parliamentary elections were held now?' I recoded the answers into five categories: (1) Fidesz 20 voters; (2) Jobbik 21 voters; (3) Momentum 22 voters; (4) voters for another opposition party 23 ; (5) do not know, do not want to vote.
All analyses were carried out in the R environment. 24 6. Results

Willingness to protest participation and its perceived risks
Hungarian university students have positive attitudes towards political protest participation. 38 per cent would participate in demonstrations if they were dissatisfied with a measure of the government (5 and 6 on the six-point scale), which indicates relatively strong political activity. 25 20 The right-wing populist Fidesz has been in power since 2010. 21 Jobbik is a nationalist, radical right-wing party that gained 19 per cent of votes in the last national elections in 2018. Since 2014, the party image has been changed, and Jobbik tries to work as a more moderate center-right party. 22 The Momentum Movement is a recently founded centrist liberal party. Momentum is one of the most popular parties with university students. 23 Other opposition parties are the leftist, liberal and green parties: MSZP, DK, and LMP. 24 R Core Team (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/. 25 Compared to the whole population this proportion seems quite high. According to the representative survey, conducted by the Peripato Research Group, in 2014, only 12 per cent of the people over age 18 would be willing to demonstrate against the government. Not only willingness is higher among university students but actual political protest activity as well. Nineteen per cent of students have participated in protest over the last 12 months, while this proportion was 3.2 per cent in the whole population in 2014. Thus, university students are more willing to protest, and they did in a higher proportion than the whole population. These results are in accordance with the literature on biographical availability (McAdam, 1986;Beyerlein and Hipp, 2006) and indicate that mobilization among university students is much easier and promises more success. If we turn to the perceived risks of political protests in Hungary, we can see that there are important differences between the four types of risk. Distributions of friends' disapproval, repercussions in university classes, counter-protesters are left-skewed, while the perceived risk of police attacks is right-skewed ( Figure 3). This means students at universities fear police attacks (and violence) the most, and they consider the three other types of risks as less likely. These differences appear in mean values as well: The least probable risk is friends' disapproval (mean=2.7), which follows repercussions at work and in university classes (mean=3.0), injury caused by counter-protesters (mean=3.5), and finally the risk of police attacks (mean=4.3). Due to the skewed distributions of items, it is worth comparing the proportions of those respondents who find demonstrations somewhat risky. These are the percentages of those whose answer was above the midpoint of the scale (5 to 7 on the seven-point scale). As we can see in Figure 3, 20 per cent of university students see the risk of peers' negative reactions, and 47 per cent see police attacks as a likely form of risk that would follow their participation (Table 1). 26 6.2 How do perceived risks relate to protest willingness?
For revealing associations between risks and willingness, I have run logistic regression models. 27 Besides the four types of perceived risks, I added gender, father's educational level, respondent's educational level, place of residence and subjective economic position as controls for socio-economic status. Also, I added political preferences as control variables in a separate model (Model II in Table 2). According to the results of the first model (Model I in Table 2), the four types of perceived risks predict willingness in different ways. The respondents who assess the risks of friends' disapproval and counter-protests as higher are also less inclined to protest. Furthermore, the perceived risk of police attacks positively predicts their willingness to demonstrate. The risk of repercussions at university, in turn, is a non-significant term in the model.
If party preferences are also controlled for in the model (Model II in Table 2), we can see that except for the risk of counter-protesters' attacks, the likelihoods of the other three risks have turned to be non-significant, which means that party preference is a strong predictor for protest willingness. Those university students who would vote for liberal or leftist opposition parties are more prone to protest than those who have no party preference, or support Fidesz. This result shows that party preferences might absorb a wide range of psychological factors (e.g. risks assessment, grievances, and dissatisfactions), political values, ideology and political identities.
INTERSECTIONS. EAST EUROPEAN JOURNAL OF SOCIETY AND POLITICS, 6(4): 63-92. I hypothesized (H1) that perceived risks positively predict willingness to participate in demonstrations. However, the findings do not support these expectations. Only one of the four types of risks predicts protest willingness positively, namely the risk of police attacks. Police attacks turned to be nonsignificant after controlling for party preferences. The other three forms of risk predict negatively or do not predict the dependent variable. Based on these results, it can be concluded that our data do not confirm the first hypothesis. Thus, it is generally not true that perceiving higher risks radicalizes university students, and makes them more prone to participate in protests.

Explaining perceived risks
In the last section, I argued that the perceived risk of friends' disapproval and of counter protesters' attack decreases, whereas the risk of police attacks increases the chances of higher protest willingness. Holding the socio-economic status constant, these factors correlate the most with the dependent variable.
In this section, I examine how socio-economic status, political attitudes, and party preferences shape perceptions of the three types of protest risks: risk of friends' disapproval, counter protesters' attacks and police attacks.
I fitted three logistic regression models to explain all three types of risks (see Table A1, Table A2 and Table A3 in the Appendix). The first model (Model I) contains only the socio-economic variables. In the second model (Model II), political attitudes are added. Finally, in the third model (Model III) I added partypreference.
The dependent variables were recoded into a dummy variable. 1 denotes that the perceived risk is higher than the midpoint (5 to 7 on the 7-point scale), otherwise, the value of the variable is 0.
Results in Table A1 show that the peer effect, friends' disapproval does not hinge on respondents' socio-economic status. There are no significant differences between social groups regarding the perceived risks of negative peer reactions. It is also independent of dissatisfaction with democracy, and the moderate-radical ideology orientation (Model II in Table A1). Nor do party preferences have a significant effect on risk perception (Model III in Table A1). The only factors that could predict a higher level of perceived risks are the liberal and rightist ideologies and political interest. In Table A1, we see that political interest increases the chances of a higher level of risk assessment by 1.52 (p=0.042). Moreover, liberals and those with rightist attitudes are more likely to have a higher level of risk. A one-unit increase (one standard deviation) on the left-right scale increases the odds to perceive a higher level of risk by 1.3 (p=0.019). Also, the conservativeliberal attitude shows a similar, but somewhat weaker association (OR=1.26, p=0.051). It seems that those of liberal or rightist political orientations are more likely to fear their peers' negative reactions than conservative and leftist students.
The interpretation of these results is not easy. In Hungary, the left-right ideological polarization is among the highest in Europe (Patkós, 2017;Vegetti, 2019), but other ideological cleavages are also deemed important in party politics INTERSECTIONS. EAST EUROPEAN JOURNAL OF SOCIETY AND POLITICS, 6(4): 63-92.
(e.g. Kostelka and Rovny, 2019). The question we asked in the survey referred to the risks of a government-critical rally, thus I would assume that those students who accept and follow the government's rightist, anti-liberal, and conservative rhetoric 28 are more prone to evaluate their friends' disapproval as a risk of their participation. The coefficients of ideological orientations in the regression model (Table A1) however show a different picture. They rather suggest that there are differences between liberal, right-wing on the one hand, and conservative, leftwing political milieus, on the other. Liberal and rightist students think that their friends would react negatively to their political activism. This means that political ties and personal networks in the liberal and right-wing milieus may shape political participation in a different way. The fact that in the third model (Model III in Table A1) party preference does not predict risk perception and ideological stances remained significant factors that bolster this approach. Party preferences do not attenuate the effect of ideologies. Therefore, the above-described differences are between political milieus and are not due to partisan polarization.
Unlike friends' disapproval, perceived risk of counter-protesters' attacks could be significantly predicted by socio-economic status. The results of the first model (Model I) are seen in Table A2. Men tend to have a higher level of risk assessment regarding counter-protesters' attacks (OR=1.40, p=0.038) than women. Moreover, among those students who come from a more affluent family background (e.g. father's educational level and subjective economic status are higher) the odds to perceive a higher level of risk are significantly lower. If we add political factors to the regression model (Model II and III in Table A2.), we can see that dissatisfaction with democracy increases the odds of higher risk assessment (OR = 2.04, p=0.001).
There are also gender differences in estimating the likelihood of police attacks. Men have higher chance to perceive higher levels of risk than women (OR=1.59, p=0. 002). Also, respondents with a higher socio-economic status (father's educational level and subjective economic status are higher, ) have a lower chance of perceiving higher levels of risks. Nevertheless, a higher level of education (MA or PhD) increases the odds of perceiving a higher level of risk of police attacks. Finally, there are no significant differences depending on participants' place of residence.
Turning to the role of political factors, we can see that dissatisfaction with democracy increases the odds to perceive higher levels of risk. Left-right and liberal-conservative ideological orientations and political interests do not have a significant effect on risk perception. However, more radical students consider police attacks as a more plausible risk of their participation (Model II in Table A2).
These predictions are stable after controlling for party preferences (Model III in Table A2); the coefficients remain almost the same and are still significant. Respondents who would vote for the governing Fidesz party show significantly lower odds to see police attacks as a plausible risk during an anti-government demonstration. Support for oppositional parties, however, does not positively or negatively predict the dependent variable (the reference category is the group without any party preference).
Results of the regression models (Model I-III in Tables A1, A2 and A3) indicate that there are important differences between the three types of risks that influence protest willingness. The risk of peers' negative reaction does not depend on socio-economic status. Liberal and rightist ideological stands increase the odds to perceive a higher level of that type of risk. On the other hand, the risk of police attacks seems to be more politicized, since those who are non-Fidesz supporters, are dissatisfied with democracy, or profess radical rather than moderate values regard police attacks as a more plausible type of risk.
It is likely that supporters of Fidesz and the Orbán-regime do not read news reports about protests, or at least they do not believe these stories and ignore the possibility of a repressive state. Otherwise, party preferences do not fully determine the level of perceived risks, since ideological orientations, interest in politics and socio-economic status have their own predictive power.
Based on the regression models, I cannot confirm all the hypotheses I formulated (H2.1-H2.4). Regarding the risk of friends' disapproval, neither socioeconomic status nor party preferences seem to have a significant effect on it. Therefore, hypotheses H2.1, H2.2 and H2.4 are rejected. On the other hand, the hypothesis regarding the effect of political attitudes (H2.3) seems valid.
Although I do find gender differences in physical risk assessment, these do not point in the expected direction. It is men who have a higher chance to perceive higher levels of physical risks rather than women. This difference to earlier studies could be explained with the special character of our sample (young, Hungarian university students). However, because of the lack of other surveys, I cannot compare these results to other samples of university students, or to representative samples of the Hungarian population.
I find that socio-economic status, political attitudes and party preferences significantly predict physical risk perception (the risk of both counter protesters' and police attacks). Lower social status (H2.1) and dissatisfaction with democracy positively predict the likelihood of physical risks of protest (H2.3). Moreover, Fidesz supporters seem to perceive police attacks as less likely (H2.4). Thus, in the case of physical risks, the regression models using the Hungarian student sample lend support to three of the four hypotheses.
In the 'Hypotheses' section, I suggested that risk assessment may purely reflect party preferences and ideological orientations. However, the regression analyses above clearly show that perceived risks of protest participation do not work as a simple proxy for political preferences. They are politicized, but in different ways and at different levels.

Discussions and conclusion
This study reinforces my assumption that empirical work on collective actions should be extended to perceived risks. Findings of the Active Youth Survey project show that a significant proportion of university students regard demonstrating against the government as a risky political action. 20 per cent of our respondents see the risk of peers' negative reactions, and 47 per cent see police attacks as a likely form of risk that would follow their participation. Fear of repercussions at work or at school and of counter demonstrators' attack are between these two extremes (25 and 30 per cent, respectively). These young adults consider that protest participation can have some negative consequences on their personal life.
Regression models show that perceived risks predict protest willingness in a complex way. The non-physical risk of friends' disapproval and the physical risk of counter protesters' attacks predict protest willingness negatively, but police attacks predict it positively. According to the Opp theorem (Opp and Roehl, 1990), perceived risks impose their effect through either deterrence or radicalization mechanisms. The net outcome depends on the balance of the direct negative and indirect positive causal paths. Thus, in the case of non-physical risk, the negative deterrence mechanism outweighs the positive radicalization effect. On the other hand, regarding the perceived physical risks, the indirect radicalization mechanism seems stronger.
Regression models predicting different types of risks display that the perceived risk of friends' disapproval is not shaped by socio-demographic background, however students with a lower socio-economic status see significantly higher levels of both types of physical risks.
Statistical models also reflect that police attacks are the most politicized form of risk. In other words, both party preferences and political attitudes predict significantly the perceived risk of police attacks. My non-systematic news content overview has also demonstrated that there are numerous reports about police attacks, house searches and incarcerations. On the other hand, I did not find any stories about friends' negative reactions. Thus, physical risks are much more widely discussed in the public sphere than personal conflicts or disapproval of friends and relatives. Since the risk of police attacks is more politicized and publicly discussed, it may trigger anger, grievance, or other emotions and political attitudes which, in turn, evoke political activism. As friends' disapproval has not been interpreted in politics, it does not trigger any other political factor, and remains part of people's personal life.
As described above, the Hungarian state has been centralizing power, and is successfully squeezing out civil society organizations and opposition parties from decision-making processes. All the findings of my study broaden our knowledge about Hungarian illiberalism, since participating in demonstrations is not seen as risk-free. Thus, it appears that attending protests or government-critical political actions need more cautious decisions.
Further analyses aiming to explain risk perception in the context of protest participation will have to take into consideration the features of respondents' political milieu, and their psychological setup. Future work should also consider whether the content of protests and demonstrations reaches people and what citizens' reactions to these stimuli are.