The bulk of research on contentious politics in authoritarian states has been preoccupied with harsh repression and transgressive contention (such as pro-democracy movements). Since transgressive contention challenges the regime, its institutions and laws all at once (Tilly and Tarrow Reference Tilly and Tarrow2007), these studies suggest that contentious politics in illiberal contexts is characterized by a zero-sum game between government and opposition (e.g. Johnston Reference Johnston2011; McAdam et al. Reference McAdam, Tarrow and Tilly2001), that is, either the regime is overthrown or the protest is stifled.
Nevertheless, a growing literature has found changes occurring in some non-democratic states in terms of handling opposition. Over the past several decades, some authoritarian regimes have typically allowed limited forms of opposition and have adopted measures to co-opt dissent (Abdullah Reference Abdullah2017; Brancati Reference Brancati2014). At the same time, disgruntled citizens self-limit their challenges and do not seek to overthrow the regime.
Then, can the concept of zero-sum games adequately represent the character of popular contention in all non-democratic states? Existing literature, which centres primarily on the zero-sum contest between authorities and challengers, on government measures to prevent the occurrence of street protests, or on case studies, leaves this question unanswered.
To address this issue, this study uses the case of China to formulate a systematic analysis of the main pattern of state responses to protests and the dominant forms of social protests.Footnote 1 Based on an original data set composed of 1,418 protest events that occurred across China between 2001 and 2012, my statistical analysis shows that authorities tolerated the majority of protest events and that most protest events did not take transgressive forms – staying away from violence, from radical political claims (for example, opposing Communist rule), and from linking organizationally with other protests. These findings, without rejecting the repressiveness of illiberal states, provide quantitative evidence to challenge the common view that contention is often a zero-sum contest in a non-democratic context. In what follows, before introducing the data set and regression analysis, I first look at current scholarship on authoritarian contentious politics and then examine factors affecting government repression.
Contentious Politics in Authoritarian Regimes
A constricted environment of political contention characterized by repression and transgressive resistance is found in non-democratic states (Davenport Reference Davenport1995; Goldstone and Tilly Reference Goldstone and Tilly2001; Gurr Reference Gurr1986; Tilly and Tarrow Reference Tilly and Tarrow2007). Coercive capacity is essential to authoritarian stability (Levitsky and Way Reference Levitsky and Way2010: 57), and even after a great loss of legitimacy has occurred, a state can remain invulnerable to internal mass-based revolts if ‘its coercive organizations remain coherent and effective’ (Skocpol Reference Skocpol1979: 32). During the Arab uprisings in the early 2010s, the coercive apparatus proved paramount in determining the durability of authoritarian regimes (Bellin Reference Bellin2012). In this sense, popular resistance in non-democratic settings is often seen as a zero-sum game: either the authoritarian regime collapses, or the protest is subdued (Koopmans Reference Koopmans2004: 39). This is the theme of most research on contentious politics in undemocratic regimes (e.g. Almeida Reference Almeida2003; Beissinger Reference Beissinger2002; Chang and Vitale Reference Chang and Vitale2013; Grodsky Reference Grodsky2017; McAdam et al. Reference McAdam, Tarrow and Tilly2001; cf. Yagci Reference Yagci2017).
On the flip side, recent research has observed that less brutal forms of authoritarian states have emerged in recent decades. Incumbents in these regimes avoid using repression as the main means to counteract challengers, resulting in a reasonable space for political activism (Brumberg Reference Brumberg2002). Meanwhile, contentious actors, by limiting their challenges, have avoided direct confrontation with incumbents (Beinin and Vairel Reference Beinin and Vairel2011: 14–15). The Moroccan regime, for instance, showed an increasing tolerance for public expression of discontent during the two decades before the ‘Arab Spring’, when many political participants transformed their practices and programmes from street activism to cooperation with the authorities (Lust-Okar 2005; Vairel Reference Vairel2011). These studies have complicated our understanding of authoritarian politics by revealing the cunning side of authoritarian rules, yet they typically focus on measures taken by incumbents to co-opt dissent and forestall street protests, rather than on regime responses when protests have taken place.
In fact, the repressiveness of a government, no matter its regime type, is always selective by types of collective action and actors (Tilly Reference Tilly1978: 106). Even in South Korea’s highly repressive period in the 1970s, the state responded with discrimination to protests of varied characters (Chang and Vitale Reference Chang and Vitale2013). No matter how great the repressive power of an authoritarian regime, it cannot control too many people simply by threatening to use force (Magaloni Reference Magaloni2006: 19); repression is costly and risky for political elites (Acemoglu and Robinson Reference Acemoglu and Robinson2006). While repression sometimes deters dissent (e.g. Hibbs Reference Hibbs1973: 82–93) and decreases rates of protest (e.g. Olzak et al. Reference Olzak, Beasley and Olivier2003), it can also backfire and increase dissent (e.g. Almeida Reference Almeida2003; Francisco 1995), facilitate the formation of alliances between challengers (Chang Reference Chang2015) or produce more persistent challenges (Franklin Reference Franklin2015).
Take China, a strong authoritarian regime, as an example. The country abounds with social unrest today. In 2011, on average, 500 ‘mass incidents’ – an official term for popular protests – took place per day.Footnote 2 The Chinese state’s reactions to contention reveal a mixed picture. Indeed, the state has adopted various measures to strengthen its control of protest. It has deliberately increased its coercive capacity by dramatically expanding its domestic security apparatus (Guo Reference Guo2012; Wang and Minzner 2015), emphasized government officials’ responsibility for reducing the occurrence of social protests (Chen Reference Chen2012; Li Reference Li2013), censored discussions that have the potential to generate collective action (King et al. Reference King, Pan and Roberts2013) and suppressed dissent such as pro-democracy movements.
Yet still many protests are tolerated, induce government concessions, or end with a combination of repression and concession (Cai Reference Cai2010; Chen Reference Chen2012; Hurst 2008; Li Reference Li2013; O’Brien and Li Reference O’Brien and Li2006; Su and He Reference Su and He2010). The regime has even encouraged narrowly targeted protests to identify social grievances, to monitor lower levels of government and to remedy the weakness of its political system (Chen Reference Chen2012; Dimitrov Reference Dimitrov2008; Lorentzen Reference Lorentzen2013). In many cases, officials manage conflicts through ‘protest bargaining’, namely, a market-like exchange of compliance for benefits (Lee and Zhang Reference Lee and Zhang2013). Nowadays, police have shifted from a strategy of deterrence and quick suppression to a more permissive strategy of containment and management (Tanner Reference Tanner2004).
As for Chinese protesters, they sometimes take to the streets and create public disruption while, at the same time, they strive to limit their activities (Perry Reference Perry2010) and to keep a balance between defiance and obedience (Chen Reference Chen2012). Usually, they prefer to keep their confrontation and violence strictly between themselves and specific local officials while affirming their recognition of the legitimacy of central authorities and the larger political system (O’Brien and Li Reference O’Brien and Li2006). Many protest actions are ‘boundary-spanning’ (neither clearly transgressive nor clearly contained), combining lawful tactics with disruptive but not quite unlawful actions (O’Brien Reference O’Brien2003) or conjoining legal challenges to injustice with extra-legal forms of civil disobedience and collective protest (Zweig Reference Zweig2010).
The burgeoning scholarship thus has suggested that the notion of zero-sum games seems inadequate to represent the whole landscape of contentious politics, at least in some undemocratic states. To what extent is this true? This article uses the case of China to answer this question. While China does not stand for all types of authoritarian regime, the multifaceted nature of contentious politics in the country provides an opportunity to analyse the varieties of state responses to challenges and to explore an array of protest tactics and goals. Earlier scholarship has provided an informative depiction of contention in China, but it does not examine to what extent the above-discussed characters of protests are typical in the country. Most research has been case studies and its biggest concern has been why some protests succeed while others fail (e.g. Mertha 2008). A few works have taken valuable steps in generating data sets on Chinese protests (Cai Reference Cai2010; Chen Reference Chen2012; Tong and Lei Reference Tong and Lei2013), yet these data sets, not including certain critical types of contention, are not suitable for identifying the lines between toleration and repression. Further, although a handful of studies have proposed what kinds of protests are repressed and what kinds are tolerated (Reny and Hurst Reference Reny and Hurst2013; Selden and Perry Reference Selden and Perry2010; Zhao Reference Zhao2010), they have not followed up with quantitative testing. Presenting a systematic analysis of patterns of repression and of types of protests is critical because it contributes to a more comprehensive knowledge of contentious politics in illiberal states. The following sections will present a broad picture of state repression and social protests in China.
Disaggregating Repression and Protest
To investigate state repression, this article centres on protest policing, or the police handling of protest events (Della Porta and Reiter Reference Della Porta and Reiter1998). Authorities may employ a variety of repressive strategies, including coercion, channelling (for example, tax restrictions on targeted groups) and overt and covert repressive actions (Earl Reference Earl2011; Levitsky and Way Reference Levitsky and Way2002). Policing, of course, is merely one of a wide array of repressive strategies, but it represents the most public and one of the most common forms of repression (Earl et al. Reference Earl, Soule and McCarthy2003). Studying protest policing is critical for enhancing understanding of the relationship between protests and the state, since police embody the objectives of the broader political-economic elite (Davenport et al. Reference Davenport, Soule and Armstrong2011) and can be ‘conceived as “street-level bureaucrats” who “represent” government to people’ (Lipsky Reference Lipsky1970: 1).
A large body of literature has addressed the question of how and why authorities respond to protests in the way they do. Among the elements found in this literature, two factors stand out: the character of protest and the power of social groups.
The Character of Protest and Repression
Claims
Prior studies of liberal democracies have found that authorities are more likely to quell challengers who wish to displace current political leaders, dismantle the political-economic system or pursue other radical goals (Gamson Reference Gamson1990; McAdam Reference McAdam1982). Likewise, in China, protests that pursue radical goals also tend to be the target of crackdowns. Nonetheless, what is considered ‘radical’ in China differs critically from what is regarded as ‘radical’ in democracies. Claims for democracy and efforts to contest the Communist Party’s monopoly on political power are prohibited or strictly constrained. On the other hand, since the reform era began in 1978, the Chinese state has become more willing to tolerate protests addressing economic, environmental and even anti-corruption issues (Selden and Perry Reference Selden and Perry2010). Divisions between different levels of government sometimes make official concessions possible (Cai Reference Cai2010; Chen Reference Chen2012; Tong and Lei Reference Tong and Lei2013). For instance, the gap between policymaking by the central government and policy implementation by the local government may empower protesters to legitimate their claims and contentious actions (O’Brien and Li Reference O’Brien and Li2006). In fact, such protests may help the central government to collect information regarding cases of power abuse and corruption by local cadres of which the central government might otherwise not be aware (Lorentzen Reference Lorentzen2013; Luehrmann Reference Luehrmann2003). Therefore, the central government may be more lenient to these kinds of protests than to those contentious political activities that directly challenge its rule (Reny and Hurst Reference Reny and Hurst2013; Zhao Reference Zhao2010).
Hypothesis 1: Economic, environmental, and moderate political claims are less likely to draw a police response, whereas radical political claims are much more likely to suffer repression.
Actions
In democratic regimes, research has elaborated that confrontational and violent actions often generate repression (Earl et al. Reference Earl, Soule and McCarthy2003; McAdam Reference McAdam1982). In China, disruptive actions, such as road blockades and strikes, may become weapons of resisters (Selden and Perry Reference Selden and Perry2010). Though some disruptive protests are repressed or end with a combination of repression and concession, others are tolerated (Cai Reference Cai2010; Chen Reference Chen2012; Su and He Reference Su and He2010).
Hypothesis 2: Peaceful actions (disruptive or non-disruptive) are less likely to suffer police coercion, whereas violence generally prompts repressive policing.
Organization
In democratic states, protests with more social movement organization involvement are considered stronger than the ones with less or no social movement organization involvement because they have access to more organization vehicles for pursuing grievances (Earl et al. Reference Earl, Soule and McCarthy2003). In China, however, the presence of autonomous formal organizations does not seem advantageous to protesters because the regime remains intolerant of organized activity not controlled by some branch of the Party (Bovingdon Reference Bovingdon2002). One case in point is the crackdown on attempts to form independent labour organizations (Saich Reference Saich2000). Moreover, since the state is well aware of the dangers inherent in cross-class, cross-regional or cross-nationality associations, it has usually acted swiftly and severely to crush attempts to forge such bonds (Selden and Perry Reference Selden and Perry2010; Tanner Reference Tanner2004) but has shown considerable leniency towards conflicts that were more homogeneous in composition and locale (Perry Reference Perry2007).
Hypothesis 3: Protests linked organizationally to other protests increase the likelihood of protest policing compared with those not linked.
Hypothesis 4: Protests organized in an informal way reduce the probability of protest policing compared with those organized in a formal way.
Protest size
Scale matters as well. According to previous studies, large protests are more threatening and more likely to lead to repression than small ones (Davenport et al. Reference Davenport, Soule and Armstrong2011; Earl et al. Reference Earl, Soule and McCarthy2003).
Hypothesis 5: A large protest is more likely to invite police presence and response than a small protest.
Social Groups and Repression
Governments also respond selectively to different kinds of groups. Authorities can be seen as opportunists that will repress when they believe they can win (Gamson Reference Gamson1990). Thus, subordinate groups – such as racial and ethnic minorities – are subjected to harsher repression. Recent quantitative research, for instance, has shown that African American protesters have been more likely to suffer repression in the US during some time periods (Davenport et al. Reference Davenport, Soule and Armstrong2011). By contrast, the persecution of mainstream, entrenched groups is likely to meet powerful resistance from the elite and fail to succeed in stifling dissent (Goldstein Reference Goldstein1978). By and large, the more powerful the group, the less repression it receives (Tilly Reference Tilly1978). On the flip side, Charles Tilly (Reference Tilly1978) also emphasizes that groups with a little power suffer more repression than do the completely powerless because the latter pose no threat to the government and their small-scale collective actions are too weak to bother with.
In China, peasants and ethnic minorities are conceived of as two important subordinate groups.Footnote 3 Firstly, the gap between urban and rural is exceptionally large. The average income of rural residents is far less than that of urban residents.Footnote 4 Collective resistance such as land protests staged by peasants tends to lead to state suppression (Cai Reference Cai2010). Moreover, despite affirmative-action-type policies, minorities are treated as peripheral peoples, marginal to power and politics (Bulag Reference Bulag2010). Within minorities, the state also has disparate attitudes. It is, for instance, more tolerant of Hui protests than Uyghur unrest (Gladney Reference Gladney2004).Footnote 5 Uyghur and Tibetan relations with Han Chinese (China’s ethnic majority) are considered the worst among the minorities (Mackerras Reference Mackerras2004), and the Chinese government has often been criticized for human rights abuses in Xinjiang or Tibet. On the other hand, neither peasants nor minorities are completely powerless groups. Instead, the Chinese regime has been deeply concerned with rural unrest and minority protests, especially those deemed to threaten social stability or the unity of the country (Bernstein and Lü Reference Bernstein and Lü2000; Mackerras Reference Mackerras2004).
Hypothesis 6: Protests by peasants are more likely to be policed aggressively than protests by other groups.
Hypothesis 7: Protests by ethnic minorities (especially Uyghurs and Tibetans) are more likely to be harshly repressed than protests by other groups.
Data and Methods
To date, there is no nationwide data set available to serve a systematic analysis of repression pattern and protest forms in China. The Ministry of Public Security has relevant statistics, yet they are not open to the public. As in other authoritarian regimes, social protest in China is a sensitive issue connected directly to state legitimacy and social stability. Given this lack of data, several scholars have compiled their own data sets, but each has significant limitations that make them unsuitable for identifying the boundaries between repression and toleration. They only include non-regime-threatening protests (Cai Reference Cai2010; Chen Reference Chen2012), merely comprise large-scale protests (each with over 500 participants; Tong and Lei Reference Tong and Lei2013), or only include protests taking place in a single city (Chen Reference Chen2012).
To overcome the data limitations, I generated a data set of 1,418 protest events that occurred across China from 2001 to 2012, a period when the number of protests skyrocketed. This nationwide data set consists of both political and non-political and both large and small protests. To create my data set, I adopted the method of ‘protest event analysis’ (PEA), which is used to ‘systematically map, analyse, and interpret the occurrence and properties of large numbers of protests by means of content analysis’ (Koopmans and Rucht Reference Koopmans and Rucht2002: 231). Being widely employed in studying contentious politics in diverse regions (e.g. Beissinger Reference Beissinger2002; Olzak Reference Olzak1992), PEA is a useful tool kit for the quantification of many properties of protest.
The unit of analysis of this research is the protest event. A social protest can include one event or multiple events. Following the guidance in the Dynamics of Collective Action Project, a single event in my research should satisfy all the following conditions: (1) it includes action that is mostly continuous – no gaps of more than 24 hours in time (weekend gaps are acceptable for labour and school protests); (2) it is located within the same city or the same part of the city; (3) it includes the same (or a subset of the same) participants whose goals are not different.Footnote 6
Due to state censorship, a great many protests are not reported in mainland China. Instead, overseas media serve as a better source of information. When constructing my data set, I relied on data collected by Boxun.com, which is an overseas Chinese community website founded in the US in 2000. Boxun.com provides nearly 2,000 incidents of resistance across China from January 2001 to December 2012, gathered largely from journalistic sources. Of these, 1,418 protest events have been selected according to my definition of ‘social protest’ (noted above). My data set includes protests that took place in all provincial administrations in mainland China.
Data Reliability and Biases
While my data set might be criticized in terms of reliability and potential biases, I believe that these issues do not affect my results in any major way. Firstly, apart from reprinting news from other media sources, Boxun.com takes the form of citizen journalism: citizens play an active role in collecting, reporting, analysing and disseminating news and information (Bowman and William 2003). In my data set, 1,110 protest events are reprinted news or confirmed by reports from other media channels, including the Western media (e.g. BBC and VOA) and websites of human rights groups (e.g. China Labor Bulletin and Information Center for Human Rights and Democracy), whereas the remaining 308 events do not clearly list their information sources other than Boxun.com. Reports of the 1,110 events are, indeed, more reliable, but since a number of reports of the 308 Boxun-specific events include attached photos or videos, which usually display protesters’ demands and the sites of events, this evidentiary support helps confirm the basic information of the events. I also made an effort to cross-check these Boxun-specific reports. As a robustness check, I ran regression analyses on two different sets of events: one including all 1,418 events and the other exclusively comprising the 1,110 events. Their results are almost identical.Footnote 7 Therefore, I am confident that the accuracy of my data set is acceptable for the purposes of this research.
Moreover, my data set is probably not truly representative of protests in China during the researched period because events were not selected based on random sampling but largely relied upon news reports reprinted by Boxun and on citizen journalists’ contributions. Boxun is critical of the Chinese regime and is, hence, inclined to collect information on the more repressive events in China. Therefore, it is likely that my data set is skewed towards protests that are suppressed. It is also fair to question the quality of news reports collected by Boxun. As the sources of information are mainly from protesters, these reports are likely to exaggerate the severity of repression. These potential biases, however, are not a severe problem for this research; on the contrary, they may help reinforce my argument. If, using a data set that is skewed towards the repressed events, I can still find a real protest space, then my argument should be strengthened. In addition, these biases help detect the limits of government toleration, since it is the more repressive events that define the boundaries between toleration and repression. On the flip side, it is fair to doubt that police coercion may be underreported. It is possible that police employed force after the news reports. This is a cause for concern for this analysis, but this concern is mitigated by the fact that such repression is likely to be reported later on. My data set includes a number of protest events with follow-up reports on repression.
Dependent Variables: Police Reactions
My goal is to analyse variations in repression through examining protest policing. I divide my analysis into two steps.Footnote 8 Firstly, I focus on police presence or absence. Secondly, given police presence, I divide policing strategies into four mutually exclusive categories: (1) taking no or limited action (such as erecting barricades and traffic control); (2) using violence only (e.g. grappling, beating and using tear gas); (3) making arrests only; (4) using violence and making arrests together.Footnote 9
In authoritarian regimes, it is common for local agents to employ thuggery (Johnston Reference Johnston2012; Levitsky and Way Reference Levitsky and Way2010). China is no exception. Some Chinese grassroots officials occasionally hire thugs to harass or attack protesters (Cai Reference Cai2010). Thus, this research also takes extra-legal methods of repression – hiring thugs – into account. If thugs were employed but not reportedly deployed by the third party, they are treated as part of the police violence. Thugs appeared in a total of 33 events in my data set; in 20 of these events it seems that they were dispatched by the government. This number is small and for some events both thugs and police appeared. In brief, policing became slightly more aggressive when thugs were included.Footnote 10 In the following sections, protest policing also includes thugs’ activities.
Overall, Figure 1 demonstrates that a real space for protest exists in China: 59 per cent of protest events were tolerated, that is, police either did not show up at the event or, if they were present, they did either nothing or took only limited action. This is true even though the data set may have potential biases for exaggerating the proportion of protests that were repressed (as previously explained). The other 41 per cent of protest events were met with some kind of coercive measures (see Figure 1). This result coincides with Yanqi Tong and Shaohua Lei’s (2013) statistical findings on large-scale social protests, in which the government tolerated 65.7 per cent of protest incidents.Footnote 11
Figure 1 Police Reactions: A Substantial Protest Space in China Source: Author’s data set. Note: N=1,418.
Nonetheless, the seemingly high proportion of state toleration cannot be taken at face value. It does not speak for a liberal environment in which protest activities are not risky and seldom receive repression. If most protesters confine their claims, actions and organization within the narrow range regulated by the state, even if a majority of protests are tolerated, the space for protest is still quite limited. Hence, it is crucial to examine the character of protests.
Independent Variables
Protest character. As its first measure, ‘protest claims’ are divided into economic claims (for example, pressing for higher pay), environmental claims (such as a demand to relocate a polluting factory), moderate political (for example, opposing misconduct of a specific official), radical political (such as a demand for an end of one-party rule) and a residual ‘other’ category. I create several dummy variables for each kind of claim. For instance, a dummy variable for an environmental claim is coded 1 when the predominant concern of a protest is regarding environment issues and no radical political claims are raised. The dummy variable for a radical political claim is treated as a reference group in regression analysis. ‘Protest actions’ are classified into violent actions (for example, attacks and riots), peaceful disruptive actions (such as demonstrations and strikes) and peaceful non-disruptive actions (for example, publishing protest letters online). I employ a dummy variable for each sort of action and treat violent protests as the reference group when doing regression analysis. With regard to ‘organization’ – a third measure of protest character, I use two dummy variables to describe two types of protest organization: isolated and formal organization. Isolated organization is coded 1 when no linkages across regions, workplaces or social groups are established. Formal organization is coded 1 when a protest is organized by an autonomous formal organization with a recognizable name and clear leadership. The last aspect of protest character is protest size, measured by the number of protesters reported at the event.
Social groups
I take peasants and ethnic minorities as examples to ascertain whether police treat subordinate groups discriminately. A peasant protest refers to a protest exclusively launched by peasants, with no presence of protesters from other social groups. Likewise, I define minority protests as protests staged only by ethnic minorities. A protest by ethnic minority peasants is treated as a minority protest. In my data set, among all minorities, Uyghurs and Tibetans staged the most protests (87 per cent out of 98). I create dummy variables for each of the groups mentioned above.
Control variables
The location of a protest might also have an impact on protest policing. Harsh crackdowns on a protest that occurs in a city are usually more likely to receive public attention and have a more negative effect on regime legitimacy than quelling a rural protest. Additionally, a control variable for the duration of a protest is included in the model, for authorities usually have more time and opportunity to respond to longer events.
Estimation Techniques and Results
Using the data set I compiled, I test the hypotheses already mentioned. My analysis takes two steps to assess how police (or thugs) respond to different kinds of protests. First, I examine under what conditions police are likely to show up at the event, using binary logistic regression (BLR). Next, I scrutinize multiple forms of policing strategies given police presence by analysing the model of multinomial logistic regression (MLR).
Police Presence at Protest Events
In the first step, the dependent variable, whether police attend an event or not, is a binary variable. In this model, the dependent variable is the log odds of police presence at a given protest event. To ease the interpretation of BLR results, I present the odds ratios, which represent the odds of observing police presence at a given event versus not observing its presence.
Results
Table 1 reports a series of BLR results. In Model 1, police presence is estimated simply by measures regarding the character of protest. Most of the character measures (except informal organization) are significant and in the expected direction. These results hold after entering control variables – protest locations and duration (Model 2). Model 3 only examines the impact of social groups on policing: it indicates that both peasants and minorities increase the odds of police presence, as expected. These results remain constant after introducing control variables (Model 4). The influence of social groups, however, is weakened after incorporating character measures: the effect of minority protests becomes insignificant (Model 5). Yet the impact of peasant protests remains robust. Building on Model 5, Model 6 adds control measures. The two models show striking continuity and the impact of protest character and social groups holds in Model 6. Overall, the full model confirms most of the hypotheses regarding protest character and social groups. Below, I report the results in Model 6 in details.
Table 1 Binary Logistic Regression Models Predicting Police Presence at Protest Events in China, 2001 to 2012
Source: Author’s data set.Note: Main entries are odds ratios with standard errors in parentheses. The models above include all control variables, yet results for controls are not reported here. R1 Reference group: radical political claims; R2 Reference group: violence.
*p<0.05, **p<0.01, ***p<0.001 (two-tailed tests).
Firstly, radical political claims are more likely to prompt police attendance than other claims. Compared with the former, protests advancing economic, environmental or moderate political claims significantly reduce the odds of police presence respectively by 20, 16 and 16 per cent. In respect of actions, peaceful disruptive protests are 25 per cent and peaceful non-disruptive protests are 33 per cent less likely to draw police presence than violent protests. With regard to organization, isolated protests decrease the odds of police presence by 8 per cent compared with protests that are linked organizationally with other protests. The forms of organization, unexpectedly, have no influence on police presence. Yet a larger protest size increases the likelihood of police attendance. In addition, peasant protests are 12 per cent more likely to draw police presence than others; whereas minority protests do not predict police attendance.
As I am also interested to know whether police discriminate against Uyghurs and Tibetans, I ran regression analysis for them as well. Model 7 resembles Model 6 except that environmental and moderate political protests become non-significant. In general, Uyghurs and Tibetans, like minorities as a whole, do not significantly invite police presence, which will be analysed later. To sum up, regarding police presence, BLR analysis suggests that my hypotheses regarding protest character are mostly supported.
Police Actions at Protest Events
What do police do, once they are present at protest events? Simply showing up without coercive response is still a relatively tolerant response. Therefore, it is necessary to distinguish the types of police reaction to protests. I divide protest actions into four categories, as listed above. Here I use multinomial logistic regression models (MLR) to estimate coefficients for the impact of explanatory variables on particular outcomes of the dependent variable, relative to a baseline outcome (Long and Freese Reference Long and Freese2006). Like the BLR analysis, I also present the odds ratios for the MLR model. In my analysis, police doing nothing or taking limited action is treated as the baseline.
Results
The MLR models in Table 2 illustrate consistent results to the BLR models. Models 8 and 9 lend strong support to a series of hypotheses concerning protest character. I focus on reporting results in Model 9 with a comparison of Model 8. First, radical political claims are more likely to result in police coercion than other claims. Specifically, economic, environmental and moderate political protests decrease the probability of arrests alone by 82, 90 and 75 per cent respectively, compared with police taking no or limited action. Economic protests also reduce the odds of arrests in conjunction with police violence by 78 per cent, compared with police taking no or limited action. Violent protests, as expected, increase the probability of coercive reactions by police compared with no or limited action. When protesters take peaceful disruptive actions, police are 84 per cent less likely to use violence alone and 84 per cent less likely to use violence in combination with arrests than they are to take no or limited action. Other variables, such as organization and ethnic minorities, surprisingly, have no significant effect on police actions. In brief, advancing radical political goals and taking violent actions in protests are triggers for police coercive response.
Table 2 Multinomial Logistic Regression Model Predicting Police Reactions at Protest Events in China, When Police Were Present, 2001 to 2012
Source: Author’s data set.Note: Main entries are exponentiated MLR coefficients with standard errors in parentheses. The models above include all control variables, yet results for controls are not reported here. R1 Reference group: radical political claims; R2 Reference group: violence. Police doing nothing or taking limited action is the reference category.
*p< 0.05, **p<0.01, ***p<0.001 (two-tailed tests).
Since I am also interested in studying police reaction to Uyghurs and Tibetans, I ran an MLR analysis for them as well. Results of Model 10 resemble Model 9 except that the impact of moderate political protests on arrests alone becomes insignificant. Therefore, Uyghurs and Tibetans, like minorities as a whole, have no influence on police coercion (to be explained below).
Explaining Unexpected Results
From the regression analysis, I find several unexpected results.Footnote 12 Among them, the effects of minority groups and formal organization deserve special attention. Why do regression results find no support for the hypothesis that Uyghurs and Tibetans receive harsher punishment? This is because Uyghurs and Tibetans engage in protests that have an exceedingly high proportion of radical political claims. When controlling for protest claims, the effect of minority protests becomes insignificant.Footnote 13 These results, however, do not suggest that Uyghurs and Tibetans are not singled out for repression. Uyghur and Tibetan protests are usually interpreted by protesters and authorities through an ethnicity lens (Hastings Reference Hastings2005), implying separatist claims. In contrast, ethnicity and separatism are not an issue among Han Chinese. When separatism is regarded as off limits by the state, the protest space for Uyghurs and Tibetans is much narrower than for Hans.
Surprisingly, protests with formal organizations do not predict protest policing. This is at odds with the rigid legal restraints on freedom of association in China. I speculate several scenarios to explain this contradiction. Protesters may have had a formal organization, but police may not have been aware of it and thus did not respond. Or perhaps no formal organization actually existed, but the police imagined its existence and used repression. Both scenarios can blur the boundaries between repression and toleration and thus contribute to the insignificant result of formal organization. My data show that only a tiny proportion (4 per cent) of protest events reportedly involved formal organization, which provides two insights. Firstly, protesters might have established formal autonomous organizations, yet for fear of repression did not disclose such information in the news reports. Secondly, the strict constraint and severe crackdown on independent associations might have deterred protesters from establishing formal organizations. Both situations suggest the state’s repression of formal organizations.
A Non-zero-sum Game
The regression results can be illustrated on a continuum of toleration and repression. As shown in Figure 2, on the left end of the continuum, protests raising economic, environmental and moderate political claims, and taking peaceful disruptive actions increase the odds of toleration; on the right end, protests advancing radical political claims, taking violent actions and linked organizationally with other protests predict repression.
Figure 2 Mapping the Space for Protest in China Source: Author’s data set.
The continuum in Figure 2 assists in locating transgressive protests, which should be featured on the right end of the continuum. In my data set, only 5 per cent of protest events put forth radical political claims, 20 per cent use violence and 40 per cent have links with other protests. It is noteworthy that my data set may disproportionately include protests with radical political claims, because Boxun disapproves of the Communist regime and is inclined to collect reports of protests with radical political claims. Even so, only 5 per cent of events fall into this category. This result is consistent with the analysis by Jianrong Yu (Reference Yu2008), who estimates that 90 per cent of mass incidents in China are rights protection actions and do not challenge the political power. All in all, my findings indicate that non-transgressive protests are the dominant form of contention in China today.
As discussed earlier, the statistical analysis reveals a real space for protest existing in China: nearly two-thirds of protest events were tolerated by authorities. This is true even though the data set may have potential biases for exaggerating the proportion of protests that are repressed. Taken together, we can conclude that popular contention in China is featured by a non-zero-sum game: the state permits some (albeit limited) space for protest and most dissidents restrict themselves to this space. This is at odds with the established wisdom of contentious politics in authoritarian regimes, which expects a prevalence of transgression and repression in such countries.
Conclusion
Through extensive event data, the findings of this article provide systematic quantitative evidence that challenges the common notion that contentious politics in an authoritarian state is characterized by a zero-sum game. While some studies have recognized that even undemocratic regimes can exhibit multiple responses to protests and self-limiting resisters can create some room for manoeuvre, these findings usually rely on case studies or less comprehensive data. Hence, we know little of a broad picture of contentious exchange between government and protesters in such states. Using the case of China, this research reveals that a non-zero-sum game played by authorities and dissenters can, in fact, be a key feature in an undemocratic state.
While this article examines the existence of a protest space in contemporary China, it is important to remember that this space is limited in important ways, as grassroots claims to citizenship rights, to freedom of association and to challenge the one-party’s monopoly of power all fall into the forbidden zones. Since of all social groups, only Uyghurs and Tibetans demand regional independence or autonomy in mainland China, which is regarded as transgressive by the regime, the protest space for Uyghurs and Tibetans is even narrower than it is for other ethnic groups. In fact, that aggrieved citizens have, in most cases, confined their claim-making to the narrow space for contention reflects the state’s ability to set limits for contentious stances, means and mobilization modes.
Overall, to enrich our knowledge of authoritarian politics, this research argues that repression and radicalization should not be the only lens through which we observe these countries. While brutal repression is frequently seen in non-democratic states, we need to pay more attention to how such states also employ tolerance and more subtle forms of coercion in dealing with contention. Focusing primarily on the iron fist can overlook the Janus-faced nature of authoritarian politics. Furthermore, the article calls for a more nuanced understanding of opposition in undemocratic states. Apart from outright confrontation, challengers in many cases may impose self-censorship and engage with authorities. Thus, this study hopes to contribute to a more comprehensive and complex understanding of contentious politics in illiberal states. Future research may distinguish types of authoritarian regimes and examine how the patterns of repression and protest may differ or be similar among the subtypes of authoritarianism.
ACKNOWLEDGEMENTS
I am grateful to Joel Andreas, Lingxin Hao, Charlie Mitchell, Daniel Pasciuti, Stephen Plank, Mildred A. Schwartz and Burak Gürel for insightful comments on earlier drafts and Sahan Savas Karatasli for valuable advice regarding coding. Previous versions of the article were presented at Stanford PACS Junior Scholars Forum at Stanford University, American Sociological Association Annual Meeting, Annual Conference of the Association for Asian Studies, North American Chinese Sociologists Association Annual Meeting, Penn Symposium on Contemporary China at the University of Pennsylvania and the Asia Conference at Johns Hopkins University. I wish to thank discussants and participants for helpful comments and critiques.
