Sticks and stones : the causes of collective violence in protest movements

(1)

UNIVERSITEIT VAN AMSTERDAM

Sticks and Stones: the

causes of collective

violence in protest

movements

Master Thesis

Political Science

Ludovico Alcorta (10599649) 7/4/2014

The analysis concludes that the dynamics between the state and the campaign are the main determinants of political campaign violence, and are hardly mediated by interactions. The political structure of the campaign is also found to have a significant effect on the type of collective action used by the campaign. This research analyzes the causes of political campaign violence. It identifies the characteristics of states, campaigns and external actors and examines the extent to which they cause a campaign to choose for violent or nonviolent methods of resistance.

(2)

1

Foreword

There is plenty of gratitude to go around. First of all I would like to thank my parents, who have always had faith in me. I would like to thank Dorkhanay for supporting me throughout and pushing me to improve. I would like to express my gratitude to my supervisors, Burgoon and Underhill, for all their input and guidance in the research process. A note of appreciation goes towards my friends for withholding from tempting me with beers (for the most part) until after the deadline. Furthermore I must acknowledge FIFA for holding an amazing World Cup during crunch time and making it a challenge for me to stay focused on writing. Finally, I want to thank my namesake, Ludovico Einaudi, for being my musical companion whilst writing this dissertation, keeping me awake during late hours and inspiring me to reach new heights. This would not have been accomplished without you all.

(3)

2

3.2.3 Control variables ... 31 3.2.4 Interaction variables ... 31 3.3 Data analysis ... 33 Chapter 4: Results ... 35 4.1 Basic Variables ... 35 4.1.1 State-campaign dynamics ... 35 4.1.2 Campaign ... 37 4.1.3 External actors ... 39 4.2 Interaction Variables ... 41 4.2.1 State-campaign dynamics ... 42 4.2.2 Campaign ... 42 4.2.4 External actors ... 43 Chapter 5: Conclusion ... 46 References ... 51 Databases: ... 54 Appendix ... 55

(4)

3

Chapter 1: Introduction

A mainstay of democracy is the premise that individuals in society are allowed to express their political discontent when they feel their country is headed in the wrong direction. Generally this is done within the electoral system, through voting, but if the formal political system is not functioning well or fully, then individuals may resort to protest to express their dissatisfaction and political demands. Protesting is a quintessential and frequently occurring phenomenon in contemporary politics, and how governments do and should deal with it is the focus of extensive political science literature. An unfortunate but important phenomenon that can change the nature and success of protests is the presence of violence. Some recent examples of such phenomena are the rioting in Ukraine, Thailand and Venezuela. While a great many protests remain within the realm of nonviolent if irregular politics, in some cases countries have experienced peaceful protests that over time have turned violent. Most of the relevant literature focuses on the causes of protest or examines the onset of civil war, but the causes of violence within the context of protests remains understudied. This research seeks to contribute to this field by analyzing why it is that some protests but not others turn violent.

In their seminal work, Chenoweth and Stephan (2011) posit that nonviolent campaigns are far more successful than violent ones. Arreguín-Toft (2005) considers violence or nonviolence to be a choice within a kind of prisoner’s dilemma game – if faced with a conventional strategy (violent), the weaker force must choose an unconventional strategy (nonviolent) in order to win, or vice versa. Asal, Legault, Szekely, and Wilkenfeld (2013) describe the use of violence as a choice of tactics that many organizations can and do change frequently according to their goals, their ideologies and the environment they operate in. But if we assume that it is a choice, and the most effective strategy tends to be nonviolence, why are violent approaches still used?

Violence can also follow from an irrational reaction which occurs due to an emotional provocation by another party. Sasley (2011) invites us to consider the effect of social emotions on a collective level with regards to the onset of violence. A ‘social’ emotion is one that is shared within a group. A sense of injustice is felt if another member of an individual’s group suffers under repression, which might lead to the group acting violently in response. Important to consider is how contagious or not this social emotion may be. Without spreading the proper motivation to others (through leadership for example) violence on the individual level could remain contained, since emotions are inherently on the individual level. Lyall and Hazlett’s abundant research in this field indicate that reactions to violence are not uniform and violence does not lead people to be more motivated for conflict. Apart from intergroup conflict, emotions may also have an effect on intragroup conflict. Another author,

(5)

4 de Gaay Fortman (2005) contemplates that aggression originates out of a sense of frustration, and that violence is a manifestation of powerlessness. The disadvantage with studying emotional causes of violence is that observations must be gathered on an individual level, which works for cross-sectional case studies but would make analysis spanning multiple years and countries almost impossible due to the amount of data collection that would be required.

Research question

The motivation for concentrating on collective political violence during social protests is due to recent events that have brought this issue into the spotlight. The 2014 protests in Thailand and Venezuela both started out peacefully1_{, yet at one point the outbreak of violence was triggered.}

Particularly in the case of the Ukraine, the news media treated this escalation as if it was inevitable and that the first deaths were only a matter of time, which begged the question – were the conditions and the mindset of those involved such that this path to violence was unavoidable? The research question of this thesis is: “Why does collective violence occur during protest movements?” The aim is to answer this by investigating the potential causes of violence during periods of protest using statistical analysis on large-N time-series datasets. This kind of data accounts for the time and location of the protests, and allows us to find whether the causal relationships are significant and if the conclusions are generalizable. Chenoweth and Stephan (2011) have developed such a database (NAVCO 2.0) in order to study the outcomes of nonviolence and violence during civil unrest, and this dataset shall be used for the purposes of this research. The method used to determine the causes of violence is to look at when protests turn violent and observe which are the characteristics and the situational context of the protest in the year leading up to that turning point. Each protest campaign is different and its duration can range from between a year (the minimum unit of observation, even if the actual duration is shorter) to spanning multiple years.

The causes of political violence are diverse and may range broadly from economic (Muller, 1985) to biological (McDermott, Tingley, Cowden, Frazzetto, & Johnson, 2009). The level of analysis of previous research varies from the individual to the national level. Although examining different analysis levels would provide a more complete insight of the underlying causes of violence, this would require an extensive amount of time and resources which unfortunately is not available for the purposes of this paper. The scope of this research will therefore focus solely on the collective

(6)

5 violence perpetrated by resistance movements, and in order to answer this specific question we shall concentrate the level of analysis on the political motivations for violent group actions.

We find in this research that there are three main actors at play that influence this violence: the state, the campaign and external factors such as the media or other states. These actors contain characteristics that lead to either more or less campaign violence. According to the results, the main causes of violence are a hierarchical political structure, disunity within a campaign, state repression and autonomy-seeking objectives. Numerous characteristics also point to more nonviolent methods of resistance, such as state concessions, diversity, campaign size, loose political structure and moderate campaign goals. Characteristics of external actors were not found to be significant. The effects of the interactions between variables also have a minor influence on campaign political violence.

This research will proceed as follows. The theory chapter includes a theoretical framework which will explain the concepts involved in this study and examine the existing literature on violence within the context of protests. This will allow us to gain an insight on the methods used to analyze the phenomenon, as well as explore the underlying causes that may lead to the use of violence by protest groups. We will identify the most salient roots of violence and structure them into a conceptual model. The methodology chapter explains the data that will be analyzed, the dependent and independent variables examined, and the statistical research methods used to test the hypotheses, as well as any limitations that may be present. The results chapter shall present first the findings of the logistical regressions for isolated variables followed by the logistic regressions for the interaction variables. The significant factors that lead to or away from violence will subsequently be determined. Finally, the conclusion will summarize the results, explain their meaning and discuss what their implications are for protests, states and this field of academic research.

(7)

6

Chapter 2: Theoretical Framework

2.1 Basic concepts

To begin this research, we must first understand the target of analysis and unit of observation that we wish to investigate. In order to do this we must clearly define the concepts of political violence and protest movements in the context of this research. The following section shall give a critical overview of the definitions widely used in the academic field, followed with the selection of definitions for the purpose of this research.

2.1.1 Collective Political Violence

Violence has a broad connotation, depending on who is defining it and for what purpose. The diversity of moral codes across the world and the evolving nature of our ethics create further cultural complications as to what constitutes acceptable behavior and what represents harm. A widely recognized and cited definition is established by the World Health Organization (2002):

“The intentional use of physical force or power, threatened or actual, against oneself, another person, or against a group or community, that either results in or has a high likelihood of resulting in injury, death, psychological harm, maldevelopment or deprivation.” The nature of violent acts may therefore be physical, psychological, sexual, or consist of deprivation or neglect. This definition also classifies the relationship of violence between perpetrator and victim within one of three categories: self-harm or suicide, interpersonal violence and collective violence. Since the focus of this research is on violence perpetrated by protest movements, the scope of this analysis is limited to violence by members of a group against another group of individuals, otherwise known as collective violence.

Collective violence takes on multiple forms, depending on the motives of the perpetrating agents. The WHO report (2002) further identifies three subcategories of collective violence according to the objectives sought: economic, political or social. The nature of protests is to achieve certain political goals (Chenoweth & Lewis, 2013) so for the purpose of this research we shall concentrate on the category of collective violence with political motives. It is also important to note the agents in this relationship of violence. Political violence is considered to be organized aggression that occurs within a state by a domestic population against its own government (Carey, 2006; Gurr, 1968; Moore, 1998). According to this definition, the main agents involved are the state and the opposition movement.

(8)

7 This paper seeks to discover the roots of political violence within protest movements by comparing the characteristics of violent protests with nonviolent protests. Often research that examines violent or nonviolent collective action may use definitions that overlap between these two repertoires. Since we study both types of collective action, it is important to distinguish between violent and nonviolent protests. The core difference between armed and unarmed collective action is in the use of or absence of direct (behavioral) violence (Bond, Jenkins, Taylor, & Schock, 1997). The coercive mechanisms used contrast distinctly: violent activism predominantly uses physical force whereas nonviolent movements engage in numerous activities such as public coercion or demonstrative appeals (Bond, 1994: 63). Similarly, a distinction can be drawn by other definitions used in the literature. Gurr (1970) describes violent activities as “all collective attacks within a political community against the political regime, its actors – including competing political groups as well as incumbents – or its policies”. Nonviolent resistance on the other hand can be outlined as “organized popular resistance to government authority which – either consciously or by necessity – eschews the use of weapons of modern warfare”. Another description for it is “an active and sustained collective engagement in resisting violence in all its forms (whether behavioral, structural or cultural)” (Dudouet, 2013). Sharp (1973) records as many as 198 varieties of nonviolent collective action, classified into three categories according to their strategic function: protests with the intent of expressing peaceful opposition to a policy or law, or to convince others of their standpoint; activities of non-cooperation which have the explicit purpose of defying, suppressing, restricting or ceasing existing relationships; and tactics of nonviolent intervention that involve direct physical obstructions in order to change certain situations, either positively or negatively. Scholars frequently distinguish protest campaigns as violent or nonviolent based on the prevalence of their resistance methods (Chenoweth & Stephan, 2011). Since this definition is widely recognized, simple and indicates a clear distinction between repertoires, we shall adhere to it throughout the course of this dissertation.

2.1.2 Protest movements

Protest movements have increasingly been the focus of academic attention during the latter half of the 20th_{century, as they have become a mainstay of political expression for civilians frustrated with}

their political system. One of the earliest definitions of the term is that protest is a type of behavior which can be classified as cognitive, affective and/or instrumental (Levy, 1983). Levy (1983) describes protest as an “an expression of a value, namely an attitude on what ought (or ought not) to be.” This entails that it has the purpose to initiate change, however it does not identify the type of change or the agents involved. Gamson (1975) and Tilly (1978) recognize the actors by defining protest as a form through which groups challenge the state. Bearman and Everett (1993) elaborate

(9)

8 on the motives for challenging the polity, which may vary from changing social policies to opposing the existing political framework.

Gurr (1993) classifies protest in three categories which represent increasing levels of scope and intensity: nonviolent protest varies from verbal opposition through political organizing activity to small and large demonstrations, rallies and strikes; riot varies from scattered acts of sabotage to serious and extensive rioting and armed attempts to seize power locally; and rebellion ranges from political banditry to protracted civil war in which rebel military units have areas of control. Most relevant to the scope of this paper are the first two categories which describe nonviolent and violent protest. As Tarrow (2011) sums up, the characteristics of these categories are concerned with collective political action and are implemented by organized opposition movements that consist of a subjugated minority or a disenfranchised majority. They are non-institutional in that they do not operate through the conventional political channels (Schock, 2003), and circumvent the established conflict resolution conduits procedures of the political system (Bond et al., 1997; Tilly, 1978, 2003). Additionally, Bond et al. (1997) note that a feature that pertains to both categories is that they are coercive mechanisms in that they impose or threaten to enforce negative political, economic, social or physical penalties for non-compliance.

A fundamental aspect of protest event analysis (PEA) is the choice of unit of analysis (Klandermans & Staggenborg, 2002). The issue with the previous definitions of protests is that they are focused on explaining individual events and do not take into account the complexity and duration of a protest movement. Concentrating on events makes it difficult to perceive trends and strategy, since protests are often just parts of a larger coordinated resistance movement which may include hundreds of tactics (Tarrow, 2011). This approach for analyzing collective action is outlined as a non-process approach (Olzak, 1989). It serves to summarize static relationships between levels or characteristics of units and some sort of event count. The alternative described by Olzak (1989) is the process approach. This is a more comprehensive way of observing collective action since it assesses how time and covariates, such as past events, impact the timing and order of repeatable events. Since the intention of this research is measure events over time, the process approach is regarded as the most appropriate method to use. In order to apply this approach we must observe protests within the context of a larger campaign. A campaign can be labeled as a series of observable, continuous, purposive mass tactics or events in pursuit of a political goal (Chenoweth & Lewis, 2013). This increases the scope of observation to include patterns over time and the political motivations of the movement. Henceforth we shall use Chenoweth & Lewis’ definition to conduct this research.

(10)

9

2.2 Causes of collective political violence

Collective political violence is a phenomenon which occurs due to the environment of the situation as well as the escalation of actions of the actors involved. In order to understand the nature of this type of aggression we must identify the actors involved and the relationships between them. This research focuses on violence against the state, so we examine the opposition campaign, the government, and external factors such as the media or foreign countries. Each of these actors contains characteristics which could be the causes of more or less collective violence in political campaigns. We can construct a number of expectations or hypotheses about such causes by looking at each factor in detail.

2.2.1 State-campaign dynamics

The state is an important actor in the relationship of violence. Its actions are the reason protest campaigns exist in the first place and therefore should also affect the decisions made by campaigns. If the actions of the state result in a provocation towards the resistance campaign, the campaign may subsequently resort to violence as an socially emotional reaction (Sasley, 2011). The dynamics between state and campaign can subsequently lead to an escalation of conflict if not checked. The state exercises its influence on this relationship via the several instruments at its disposal. When faced with a challenge to its rule or sovereignty, a state can decide to either work with the campaign in order to appease their demands (concessions), or against it with the purpose of quelling the movement (repression). Hafez (2003) describes these two components as political access and state repression, which are not necessarily interconnected. The state is not beholden to using just one of these options - in many cases it may alternate between these strategies, or even apply both

simultaneously. Franklin (2009) considers the absence of either approach to be a strategy unto itself, called toleration. The strategy chosen by the state is based on a cost-benefit analysis of either option or a mixed strategy in order to achieve the greatest expected overall utility (Gartner & Regan, 1996). Repression and concessions are not the only state-related factors that impact campaign violence; individual actions such as state defections may also play a role. For the purposes of this research, the focus of analysis will be on the two main factors: repression and concessions.

Concessions

Concessions are actions the state carries out in order to appease the campaign. Sometimes the state can change its behavior to accommodate the opposition without making formal or public

concessions. Chenoweth (2011) frames the action from the perspective of the opposition campaign and labels it as ‘progress’. She uses a broader range for defining accommodative policies, which extends from visible gains short of concessions to complete capitulation. This is a more

(11)

10 thought on the consequences of concessions on campaign political violence, one suggesting a positive relationship and the other implying the opposite. The majority of research thus far supports a positive correlation. Numerous scholars have advocated that accommodative policies may lead to further protest and violence (Hewitt, 1984; Wilkinson, 1999). Goldstone & Tilly(2001) subscribe to this argument, claiming that modest concessions may advertise a regime’s illegitimacy and attract new emboldened protesters to the movement. If the concessions are substantial enough however, they would be able to stop the movement. Similarly, Acemoglu and Robinson (2000) invoke the Machiavellian argument that limited concessions show weakness and may spur on future protests, and thus in order to avoid rebellion the leader should either not concede at all or make the most generous concessions possible. Focusing on terrorist organizations, De Mesquita (2005) finds that moderate factions within a movement will accept the compromises and disband, leaving more extreme factions in control. This leads to a ‘radicalization’ of the movement and ultimately to increased political violence. Value-expectancy models contend that civilians will rebel if they are certain that dissent will achieve the collective good (Finkel, Muller, & Opp, 1989; Klandermans, 1984). The diffusion of a protest to a larger group of people is influenced by the campaign’s ability to obtain concessions from the government, since a victory will convince people that their continued activism will ultimately be successful (Chong, 1991). More recently, scholars have studied the effects of concessions on violent and nonviolent protests (Rasler, 1996) and political violence (Bell,

Cingranelli, Murdie, & Caglayan, 2013). Rasler’s (1996) findings on the Iranian revolution seem to support the prior research, in that concessions increase mobilization and frequency of violent protests. The analysis yields ambiguous results though, since she determines that concessions also result in more nonviolent protests. What increases is the total number of protests, not particularly the violence. Furthermore the research only involves one case study, so the results are difficult to generalize. Bell et al. (2013) determined concession to be correlated with more political violence as civilians are allowed a platform in which to instigate politically violent acts against the state.

However he interprets ‘concessions’ as the state allowing citizens to coordinate protests rather than actually granting any demands of a campaign, which can be classified rather as the absence of repression. This interpretation is limited and does not take into account a wide range of measures a regime may undertake in order to reach a compromise with the campaign.

The view that the relationship between concessions and campaign political violence is positive appears to be counter-logical. Although protesters may be spurred on to make further demands and continue protesting until their objectives are fully met, the fact that the state is giving in to their demands should curb further violent tactics. If states decide to concede rather than repress, it is

(12)

11 precisely because they wish to curtail the use of violence on either side. Therefore concessions usually carry with them the condition that the opposition will cease hostilities. Protest movements are not necessarily beholden to these conditions, however if they are gaining political concessions from the state, they will feel more encouraged to pursue their objectives through conventional channels. Resorting to violent tactics may ultimately prove counterproductive if the government does not see the goodwill reciprocated and reverts to using repressive methods. Several scholars favour the argument that concessions lead to less political violence, although they are fewer in number. Lichbach (1987) develops a Rational Actor model which determines the effects of government actions on dissident tactics through imposing different costs of choosing a strategy. Lichbach (1987) concludes that if a government’s accommodative policies are consistent towards the tactics of a campaign, they will decrease violent dissent. These findings are corroborated by Moore (1998) when he uses sequential tests of events data to assess Lichbach’s (1987) and Rasler’s (1996) models. Franklin (2009) compares concessions with repression and toleration (doing nothing) and infers that although concessions have short-term costs linked to the resources needed to provide them and the opportunity costs of choosing a set policy, they are likely to end the current challenge. A caveat is that they may encourage future challenges in the long term by increasing campaigns’ estimates of future success. Ginkel and Smith (1999) similarly claim that concessions may deter rebellion, though at the cost of signaling weakness on behalf of the regime. Examining research on terrorism, terrorist political violence is used to extract concessions (de Figueiredo Jr & Weingast, 2001). In order for this argument to hold, there has to be credibility that terrorist threats would continue, and that it would stop if concessions are given. In a working paper, (Shadmehr) considers bystander citizens are likely to join an existing protest, but will not have the resources or

organizational skills to initiate a protest. Thus if a regime concedes to an activist’s demands, they will coopt and can end the movement as it will most likely not have the incentives to attract bystanders anymore. Grassi (2014) demonstrates in the case of Latin America that some democracies were able to reduce or eliminate political violence by offering concessions to the opposition. Concessions were also used successfully but to a far lesser extent in authoritarian regimes.

In addition, another phenomenon related to the state that has a possible influence on violence in protests is defections from the state. Case in point, McFaul (2005) includes “divisions among the regime’s coercive forces” as a key ingredient of successful colored revolutions. Both Beissinger (2007) and (D'Anieri, 2006) stress the importance of elite defection as a key factor for protest outcome within revolutions in post-communist states. The impact of defections partly depends on the presence of independent media however, since if defections must be communicated to the general public and probably will not be if the media is controlled by the state. Moreover, neither

(13)

12 research gives any insight as to the effects on political violence. On their own, defections would have only an indirect effect on campaign violence. This characteristic would perhaps have a mediating effect on other variables, so it would be interesting to study its interactions with state repertoires. The effects of concessions are complex and can incur costs in the short and long term. Although most of the research on of the consequences of government concessions points towards increasing campaign political violence, the evidence so far is inconclusive. This may be due to the fact that when analyzing the consequences of concessions, violent tactics are not always separated from contentious actions. If this differentiation is made, the results may prove the contrary – that concessions reduce violence. This research seeks to determine if this is the case, by looking at a large sample of protests spanning different countries and an extensive period of time.

H1: State concessions towards the campaign lead to a decrease in the use of primarily violent methods of resistance.

Repression

Repression is considered one of the chief instruments of the state in dealing with protest movements, and is an instinctive response when the political status quo is threatened. Repression (defined as coercion) consists of state violations of physical integrity rights (Bell et al., 2013). The effects of repression on resistance movements have been exhaustively studied in the literature. Gurr (1968), Gurr and Moore (1997), and Opp and Roehl (1990) find a correlation between repression and the onset of protest, Regan and Norton (2005) find similar results for civil war and Walsh and Piazza (2010) draw parallel conclusions with respect to terrorism. Muller (1985) compares the effects on political violence of repression with income inequality and concludes that repression has a stronger impact. He also finds the positive correlation between higher repression and more protest violence. Bell et al. (2013) concur with this view, as they determine that coercion makes citizens more willing to commit acts of political violence. Lichbach (1987) develops the hypothesis that a government’s repression of nonviolent protests will reduce the nonviolent activities of the campaign but increase its violent activities. Saxton and Benson (2006) find that repression is predominantly related to higher levels of both non-violent and violent protest, and may activate an upward spiral of violence by inciting more mobilization and rebellion, in turn leading to regime closings and more repressive authoritarian regimes. A positive relationship between repression and the onset of protest is established (O'Connell, 2008; Rasler, 1996; Siegel, 2011), as well as with repression and terrorism (Cronin, 2009; Sánchez-Cuenca & De la Calle, 2009).

(14)

13 The type of repression can also vary depending on the strategy and military or economic capacity of a regime. Repression tactics can be implemented haphazardly against all citizens of a state or it can be targeted towards certain members or elements of a campaign; often the leadership in order to decapitate movements (Cronin, 2009; Johnston, 2012). Hafez (2003) states that this distinction is crucial in analyzing the consequences of repression, as the preemptive and precise repression by the state will likely succeed in reducing political violence, while reactive and indiscriminate repression will likely incite more political violence. Even though the use of indiscriminate repression would increase the costs of protest, these costs could be spread out across a larger pool of protesters who see the increased need for their actions (Clay, 2012). Other scholars look at the effects over time - repression is likely to lead to short term pacification but more mobilization and political violence in the future (Bell et al., 2013). A note of caution is suggested by (Chenoweth & Lawrence, 2010), where they explore how inconsistencies in state repression can muddle the decision on tactics that campaigns make. In view of the literature, the consequences of repression on campaign violence are a key part of this research and can be hypothesized as follows in order to test the consensus opinion. H2: Heavier state repression will lead to higher campaign violence.

2.2.2 Campaign

The campaign is the other main actor in the relationship of political violence. Campaigns may either decide to use a strategy of violent resistance (Stephan & Chenoweth, 2008) or they may resort to violence as a socially emotional reaction to certain provocation (Sasley, 2011). Whether violence is a conscious decision or occurs due to an emotional group response, there are certain characteristics of any movement that define it, and can make it more or less predisposed to violence. We identify the most salient ones from the literature and formulate our hypotheses in accordance with the results, with the purpose of testing the existing claims on a campaign level.

Diversity

Diversity within the campaign can be a major influence in determining whether protest is peaceful or violent. A host of research covers the extent to which different types of diversity are likely to have an effect on violence in protests, such as gender (Moser & Clark, 2001; Rowbotham, 2013) and identity (Kalyvas, 2003). Gender, religious or political ideologies can determine whether violent tactics are used by organizations (Asal et al., 2013). Demographic diversity is likely to lead to less violence, as the inclusion of a more representative cross-section of society will tend to dissipate the influence of protesters who are more inclined to use violence. Academics have studied the impact of women in

(15)

14 forming organizations in which they partake. The inclusion of women in political and organizational life has a strong negative effect in terms of intrastate, interstate, and state violence (Melander, 2005). Organizations with gender-inclusive ideologies tend to be more peaceful and are far less likely to engage in violent strategies (Asal et al., 2013). Asal et al. (2013) only investigate protests in the Middle East however, which makes it difficult to generalize these results to a broader range of protests. This claim is one that we seek to test on a larger scale to see if it holds up. Ergo, we make the following claim:

H3a: Gender diversity would result in less violent resistance by the campaign.

Groups with different identities and ideologies will often ally together in order to more effectively resist the government. The presence of different identities and philosophies may frequently create tensions between rival groups. If these differences are emotionally charged and cannot be easily resolved or integrated, such as ethnic identities, they can escalate towards the use of violence. Wood (2003) claims that mobilization, violent or otherwise, is often inspired by genuine ideological commitment. Bearman and Everett (1993)contend that group identity is more important than group interest in initiating social protest. This is due to the innate nature of ethnicity and the strong emotions that are attached to belonging to a particular group identity. The existence of ideological diversity has an ambiguous effect on the likelihood of aggression, depending on whether the groups involved advocate violent tactics or not. Different ideologies will yield different choices of mobilization tactics over the course of a campaign (Gurr, 2000) and these differences will play a key factor in shaping incentives for violence (Pearlman & Cunningham, 2012). The existence of ethnic cleavages in a population can cause higher death rates from political violence, although this effect varies over time (Muller, 1985). Saxton and Benson (2006) also identify the lack of group cohesion, based on culture and race, to be an indicator of protest and violence, although he does not obtain significant results for violence. Importantly, Toft (2003) discerns that ethnic groups are far more likely to mobilize for violent collective action than ideologically based groups because their ethnic identity is more strongly rooted. The literature studied above observes the interactions between rivaling groups. We can assume however that these interactions will also reflect the dynamics of different ideologies and ethnicities within a movement, as the tensions and challenges faced are similar. In our hypothesis we will examine the effects of both ethnic and ideological diversity within a campaign on its use of violence and the differences between them to test whether the results pan out similarly to those of Toft (2003) and Pearlman & Cunningham (2012) when condensed to in-group dynamics. Therefore we will assume a positive effect on violence

(16)

15 H3b: Ethnic diversity and ideological diversity are likely to lead to higher campaign violence, with ethnic diversity having a stronger effect than ideological diversity.

Size

The size of a campaign is another integral aspect which may play a role in determining the type of resistance method of the movement. Nonviolent collective action usually requires high participation to be effective (Stephan & Chenoweth, 2008). In addition, larger campaigns can exert more pressure on a state and are therefore more likely to succeed using conventional means. Smaller groups on the other hand may consider violence to be more advantageous in accomplishing their objectives. Partly using the NAVCO 2.0 dataset, D. E. Cunningham, Gleditsch, and Salehyan (2013) found results contrary to these expectations. They do not have a convincing answer for them however, and acknowledge that larger groups are more included into political power than smaller groups. Moreover, their measure for size of campaigns is relative to the population instead of absolute, and is extracted from several other databases. Hence we sustain Chenoweth’s claim that a larger movement size indicates more peaceful protests.

H4: The larger the campaign the more likely it will rely on primarily nonviolent resistance methods.

Disunity

Resistance movements are complex phenomena which are not always recognized as single cohesive entities. To better understand the decision-making process that occurs within them, it is sometimes better to disaggregate them into the different groups/factions present in the resistance movement. As Cunningham, Bakke, and Seymour (2012) point out, when there is more than one faction in a campaign, every faction competes not only with the state, but also with each other. The competition with the state focuses mainly on public goods that benefit the group as a whole, whereas contest between factions is predominantly about power and material goods that benefit the individual faction, and is not shared with the group as a whole. Since these material goods and power are necessary for the continued survival and growth of the faction in the short run, we can assume that the internal contest is seen as more important than, or just as important as, the goal of the external contest. We can therefore expect inter-faction competition to have a profound effect on the choice that groups make within a movement between a violent and a nonviolent strategy. Each group seeks to become the central force so that they can determine the dominant repertoires (Bearman & Everett, 1993). Cunningham et al.’s (2012) research finds that more competition between factions within a resistance movement leads to higher levels of violence against the state. This is because Cunningham et al. (2013) elaborates further that internal division in campaigns make civil wars more

(17)

16 likely and nonviolent campaigns less likely, although the latter relationship is not very significant. Pearlman (2011) uses a process tracing approach to analyze the reasons for choosing violence or nonviolence in Palestine, and argues that “the path to nonviolence requires a movement to have or create internal cohesion”. Her argument explains the relationship from a different perspective but comes to the same conclusion. This paper seeks to identify the influence that disunity within a movement has on the likelihood of violence occurring. We formulate a hypothesis that follows Cunningham et al.’s (2012)

H5: The less united factions are within a campaign and the more they need to compete with each other for resources and power, the more likely they are to resort to violence.

Goals

The objective of a campaign is its raison d’être, and this characteristic can often be crucial in defining whether a violent or nonviolent approach is taken. Collective action (including violence) results from the coalition of local and central actors in pursuit of diverse objectives (Kalyvas, 2003). Behavioral shifts in a movement might further be triggered by a reevaluation of the methods and/or goals of the campaign in light of a constantly evolving environment (Dudouet, 2013). The demands of a campaign can be numerous, vary widely from policy change to complete independence from the state, and often develop over time. As one can imagine, the state would be less tolerant of challenges to their sovereignty than they would be of appeals for political reform. Gartner and Regan (1996) propose a positive linear relationship between the levels of opposition demands and the intensity of government repression towards the opposition movement. Campaigns seeking more autonomy or unequivocal secession are therefore less likely to achieve their goals through conventional means, which may lead them to turn to violence. Cunningham et al. (2013) support this argument, positing that civil war and nonviolent campaigns will be more likely when self-determination campaigns desire independence. The range of goals under analysis is limited though, as Cunningham et al’s (2013) analysis only observes independence. A more nuanced approach would be to account for a broader range of objectives and their consequences for political violence by campaigns. The hypotheses below reflect the diversity of these goals and their consequences by defining which demands will lead to more politically violent resistance methods and which ones will lead to less.

(18)

17 H6: Secession, autonomy and anti-occupation are goals that undermine the sovereignty of the state and therefore campaigns that have these demands are liable to resort to violent resistance methods to achieve these goals. Regime change, significant institutional reform and policy change can all be achieved without changing the existing political system and therefore campaigns seeking these objectives are likely to be less violent.

Structure

Finally, the organization of a campaign’s structure can influence the campaign’s propensity on whether to use violent methods or not (Pearlman, 2011). Structures can range from flat, where decisions are based on consensus, to hierarchical, where decisions are made by one or several leaders at the top and communicated down the hierarchical chain. In flatly structured movements, no one leader asserts authority over the rest and tactics are chosen either by the majority of the group, or unanimously. In hierarchical movements, leadership is instrumental in setting and changing the behavioral pattern of campaigns (Dudouet, 2013). Leaders have the ability to inspire the masses into action, though whether they use this to shift the movement’s strategy to unarmed or armed resistance depends on the values they advocate (Ogura, 2008). For example, both Hitler and Gandhi were strong and inspirational leaders, and they used their influence on their followers to pursue opposing courses of action with regards to violence. Gandhi was able to curb violent resistance against the British government for the most part, whereas Hitler incited violence against first the government and then other ethnic groups.

Although the influence of intragroup political structure on intergroup violence has been extensively studied, the exact impact it has appears to be ambiguous. On one hand, decentralized organizations are considered by some scholars to be more predisposed to conflict (Pearlman & Cunningham, 2012). Cunningham et al. (2012) show that fragmentation creates incentives for conflict by stimulating factions to impose costs on the state through violence. Another reason for this is proposed by Lawrence (2010), who argues based on a case study of countries under French colonial rule that once factions within a movement are forced to compete for resources and power, violence will increase between them. Furthermore, centralized groups may solve collective action dilemmas for initiating conflict (Petersen, 2001; Wood, 2003). On the other hand, hierarchical movements can curtail internal radical factions from intergroup violence (Pearlman, 2011; Weinstein, 2006). The topic has been especially difficult for political scientists to solve because of the endogeneity issues involved (Bearman & Everett, 1993). In the case of armed conflict, resistance movements are formed to serve military aims, and so their organizational structures are both determinants and products of conflict processes (K. Cunningham et al., 2012; De Mesquita, 2005; McLauchlin & Pearlman, 2012).

(19)

18 These findings contradict one another, and certainly do not offer a definitive answer on the nature of the relationship between intragroup political structure and intergroup violence.

A recent study has attempted to resolve this conundrum by analyzing the political structure and incidence of political violence among Native American tribes against British/U.S. forces during the American Indian Wars (Friedman, 2013). This research is interesting for several reasons. First of all it offers a relatively homogenous set of observations in very similar environments fighting the same opponent (the British army until 1776 and the U.S. army afterwards). Secondly, the extensive and systematic research on tribal history allows for a comparative analysis of violent and nonviolent tribes, as the prevalence or absence of military behavior is captured for both. Thirdly, this historical research also includes detailed insight on the political and economic structures of tribes. Finally and perhaps most importantly, most tribes were formed before there was any conflict with British or U.S. forces. This would make their political structure independent of political violence and solve the endogeneity issue that plagues scholars.

Friedman categorizes tribes into political structures by their level of centralization, according to three mechanisms: consensus building, self-policing and coordinating military action. These categories are classified in an increasing order of centralization as segmentary, band and tribe-level polities. He concludes that tribes with more centralized political structures were more likely to engage in political violence, wage large-scale “wars”, to fight more destructive conflicts, to fight for longer periods of time, and to achieve more success on the battlefield. These correlations are for the greater part substantively and statistically significant. Friedman argues that agricultural subsistence patterns can be considered as the foundations for centralized political structures in tribes. By isolating the origins of tribal political structure from conflict the agricultural subsistence pattern argument subsequently intimates a causal link between political structure and political violence. Naturally there is a considerable difference between Native American tribes and contemporary resistance movements, so this comparison must be approached with caution. Tribal political institutions were relatively non-coercive and non-hierarchical compared to resistance campaigns, let alone modern states’ military organization. Nevertheless, their political institutions are correlated consistently with conflict dynamics. If their political institutions are so limited with regards to coercive authority but still hold up against current theory, then this study can be treated as a “least-likely case” scenario in finding the empirical relationship between political structure and military behavior. Given the fact that political institutions are now more established, we might assume that this relationship has only become stronger. Another difference is the simplicity of conflict at the time compared to now. The weaponry was primitive, the military tactics were straightforward and the

(20)

19 troop sizes were small. This can be seen as an advantage however, because a simple setting permits the isolation of factors in order to directly test the dynamics between political structure and conflict. Finally, Native American tribes did not constitute a coherent entity resisting against the state. However their objectives were similar in that they were fighting to retain their lands and rights from a repressive force. In this context they could be aggregated as one movement vs. a ‘state’. The hypothesis will reflect the relationship established by Friedman in order to see if it holds up in the current context of protest campaigns and collective political violence.

H7: Campaigns with a hierarchical structure are more likely to use violent methods of resistance, whereas campaigns with a loose structure are more likely to use primarily nonviolent strategies.

2.2.3 External actors

Aside from the state and the campaign, the relationship of political violence involves external secondary actors that may influence the equation. Major actors that are recognized include the media and other states. These actors are not usually directly involved in the conflict situation but often have a strong enough influence to reframe the relationship and shape state or campaign actions indirectly.

Media

Much attention has been devoted to the impact that media has in shaping the dynamics of protest and repression because of the influence it has on domestic and international audiences. Ideally it shines an objective spotlight on the situation, and being monitored will lead actors to use less violence due to fear of condemnation or punishment if they are exposed. In reality however, media can be a complex issue because there is the possibility of bias in coverage. Disputes framed in a certain light by newspapers and television news impact the favorability of a campaign or regime (Brasted, 2005; Corrigall-Brown & Wilkes, 2011; McLeod, 1995). There is ample research that illustrates how radical movements are marginalized by media coverage (McLeod & Hertog, 1992; Shoemaker, 1984). Gamson and Wolfsfeld (1993) study the interactions between media and movement. Although they both rely on each other, movements are dependent far more on the media to broaden their scope, receive sympathy for their cause and increase mobilization. They conclude that media can steer movement strategies towards catering to mass audiences, which involves using more violent tactics since there is a higher premium on visual material. Most of the studies mentioned incorporate the effects of framing of news coverage, which can be very difficult to determine and quantify. Bell et al. (2013) looks at the effects of news coverage whilst controlling

(21)

20 for bias and reach similar conclusions. They show with statistically significant evidence that the more media coverage a state receives, the more it is prone to domestic violent protest. Thus we will assume that media coverage increases campaign violence. We will also control for reliability and accuracy of sources, hence we expect to find similar results.

H8: Higher media coverage of a campaign is likely to lead to increase anti-government violence.

Other states

Throughout history, states have meddled in each other’s affairs based on their values, affinities and interests. In the current environment where the world is highly interconnected and every action of a state is in the glare of media spotlights, it is nigh impossible for a state to react to a resistance movement without some receiving some sort of response from the international arena. The types of responses are numerous and may include expressions of (withdrawal of) support for the regime or the campaign (Tilly & Tarrow, 2006), statements of condemnation against violations of human rights (Cole, 2012), economic sanctions or financial support (Peksen, 2009), and even military intervention. These actions generally appear to be ineffective and in some cases counterproductive for human rights and political violence within the targeted country (Cole, 2012; Peksen, 2009). Repercussions have mainly been studied in terms of their impact on the state’s actions though, and only a few scholars have focused their attention on whether repercussions may influence a campaign’s method of resistance (Dudouet, 2013). Dudouet’s study considers the readiness for withdrawal of support for a movement to be a motivator for the campaign to seek for international sympathy by adopting nonviolent strategies. Walsh and Piazza (2010) look at states coping with terrorist movements. They determine that the regime’s abuse of human rights will undermine domestic and international support for the state’s counterterrorism efforts, which in turn will increase levels of terrorism. Foreign governments that advocate human rights norms may go as far as to provide material assistance to insurgencies such as in Egypt, Libya, Tunisia and Syria (Bell et al., 2013). We want to build upon this body of research and investigate whether there is a link between international repercussions against the state and a campaign’s strategy of nonviolence or violence.

H9: Economic sanctions and material repercussions are manifestations of foreign condemnation of a regime. Their effects are likely to be counterproductive and increase campaign political violence.

2.3 Conceptual model

The dynamics between all actors involved in the relationship of political violence within the context of resistance movements are illustrated in Figure 1. The three agents involved are categorized as the state, the campaign and external factors. They are tested on their impact on campaign political

(22)

21 violence - the expectation is that the state and the campaign characteristics will have a strong effect whereas external factors will have at most a minor effect. This is because structural attributes of a campaign traditionally play a role in determining the type of collective action chosen by a campaign, and the dynamics between the state and campaign are widely considered by scholars to have an influence on the (de)escalation of conflict. Conversely, the literature studying the effect of external actors provides ambiguous or insignificant conclusions. Each actor includes a number of characteristics that has a particular effect on the likelihood of violent methods of resistance being used. The most significant characteristics are outlined in the theoretical framework: repression and concessions for the state; disunity, goals, diversity, size and political structure for the campaign; and media exposure, international economic sanctions and repercussions from external actors. The state is the direct target of the movement’s demands so it can be assumed that the government’s actions of repression or concession are fundamental in defining the strategy of the movement. Structural characteristics of the campaign itself should also be fundamental in indicating the strategy it takes with regards to violence or nonviolence, although their influence may vary. As for external actors, media exposure has a direct but ambiguous effect and other states generally have an indirect influence which can affect the actions of the campaign by manipulating the state’s behavior. The reality of the dynamics of the relationship between campaigns and political violence are usually more complex than the aforementioned determinants however, since the characteristics can also interplay with each other to influence collective action strategies (Siegel, 2011). Siegel argues that to understand the efficacy of repression one must consider the conditioning effect that the campaigns objectives, the internal cohesion and the type of repression (discriminatory or haphazard) have on each other. In order to achieve a more comprehensive insight of the causal relationships, it is important to consider that the characteristics are not necessarily independent from one another and the interactions between them should be scrutinized. Therefore we create interconnections between the three categories and analyze the effects of the most significant interactions with the aim of finding significant relationships between them and the use of violent or nonviolent campaign resistance methods. The interactions will be examined to observe whether the independent causes of campaign violence are mediated when they interact with other variables.

(23)

22

Figure 1 - Conceptual Model for Campaign Political Violence

(24)

23

Chapter 3: Methodology

3.1 Data

The scale of investigation required to analyze violence in resistance movements worldwide restricts this research to empirical secondary data that has been gathered with pooled resources and its credibility verified by a consensus of scholars. Currently there are two comprehensive and publicly available datasets which cover violence in protests and campaigns on a global scale. These two datasets use an inherently different method of recording observations. Chenoweth and Stephan’s (2011) Nonviolence and Violent Campaigns and Outcomes database2_{(NAVCO 2.0) uses campaign}

years as observations, whereas the Social, Political and Economic Event Database (SPEED) is event-based. SPEED3_{only takes into account discrete events rather than analyzing the overall campaign.}

Tarrow (2011) argues that focusing solely on events hinders analyzing the patterns that might indicate a consistent strategy. Furthermore, campaigns often consist of numerous subversive activities. By looking at them individually and independently there is a risk of missing the big picture. Additionally, the SPEED dataset does not differentiate between campaign and state violence, which is a crucial distinction for the purpose of this research. It also misses out on key components that describe the context of the campaign such as diversity, structure of the campaign and external factors such as media and sanctions. It focuses rather on the specific details of the protest, which describe the minutiae of the activities that occur, yet they do not reveal the numerous underlying reasons for which the protest may have occurred. The dataset does have some advantages however, since it takes into account whether violence was advocated, what types of weapons were used, injuries rather than deaths, which provide more insight into what happens during a protest.

The researchers that created the NAVCO 2.0 database define campaigns as a series of observable, continuous purposive mass tactics or events in pursuit of a political objective. The more nuanced unit of analysis in the NAVCO 2.0 database is campaign year. This is a more comprehensive way of analyzing the campaigns than merely by events, which on their own rarely threaten the stability of governments and often do not have the purpose of overturning the system in which they occur (McAdam, Tarrow, & Tilly, 2003; McPhail, Schweingruber, & McCarthy, 1998). This definition is also more accurate than observing by country year, as numerous campaigns can be present within the same country. Categorizing campaign events by years is still not the ideal way of analyzing campaigns since one year is a long period of time in which campaign dynamics and tactics can change drastically. Campaigns can begin and end before the years is over. However recording events in shorter time periods would be an impossible task for researchers, since they cannot collect

2_{Source: See ‘Databases’ in reference list} 3_{Source: See ‘Databases’ in reference list}

(25)

24 information on the complete series of variables over the course of one month for example. This data, although more accurate in chronological terms, would not be suitable for any meaningful analysis since most of the results would be insignificant.

The NAVCO database consists of 323 violent and nonviolent campaigns throughout the world, which include 1726 observations in the period of 1900-2006. The dataset does not include all nonviolent campaigns because many of those not included as observations were not known to researchers at the time the data was being collected in 2007. Small nonviolent campaigns that were crushed in early stages may therefore be omitted, and the nonviolent campaigns that are reported will tend to be large and mature. This can bias the dataset towards success, which represents a major yet unavoidable limitation. Chenoweth & Stephan moderate this underreporting bias by comparing nonviolent campaigns with their equivalents in violent campaigns instead of examining them independently. By looking at only mature violent campaigns, the underreporting bias is balanced between both types of campaigns. Campaigns must satisfy two criteria: participation and goals. Only campaigns with maximalist objectives and with a minimum participation of 1000 people were included. The dataset was distributed among leading scholars in order to achieve a consensus sample. Additionally, each protest campaign is different and its duration can range from between a year (the minimum unit of observation, even if the actual duration is shorter) to spanning multiple years.

It is important to understand the limitations of the NAVCO 2.0 database. It was created with the purpose of identifying whether a campaign was primarily violent or nonviolent, and if it was ultimately successful in achieving its goals or not. The variables included serve more for the purpose of determining why a campaign is successful, instead of why a campaign is violent. States that set off violent resistance may differ fundamentally from those that produce nonviolent resistance. Going further, states that experience campaigns at all can be vastly dissimilar to states that do not experience a campaign. These differences may impact the chances of a campaign’s success. Chenoweth & Stephan (2011) account for the divergence between states that experience campaigns, by comparing states that are subjected to violent campaigns with states that experience nonviolent campaigns on state military strength, regime type and economic development. They further compare states in the dataset that experience any form of resistance campaign with a random sample of states that do not undergo any form of campaign. However this analysis is not

(26)

25 incorporated into the dataset itself, and therefore data on regime type, state capacity and economic development cannot be used when examining the causes of violent resistance. Additionally, other characteristics such as whether violence was advocated (by campaign leaders) and media framing are not included. These variables would have provided useful insight on what kind of leadership leads to violence and what effect media bias has on violence.

3.2 Variables

3.2.1 Dependent variable

The NAVCO 2.0 dataset has three variables that measure violence. The first variable (prim1) denotes the primary type of resistance method of a campaign throughout its duration. It remains static even though campaigns may alter their main method of resistance over the span of multiple years, and so does not take this variation into account. The second variable (prim2) identifies primary type of resistance method in a campaign year and therefore considers the nuances of deviations in tactics over the course of time. The third variable (resis_meth) gauges the (lack of) change in types of resistance methods within a campaign year, which is a more specific measurement for determining the turning point from nonviolence to violence and vice versa. However as the frequencies in Table 1 illustrate, the third variable only includes a total of 62 changes in type of resistance methods, which accounts for 3.6% of 1705 observations. This suggests that there was not much cross-year variation within campaigns. Another possibility is that they were classified as using one method for one year and using another for the next, whilst the change was not registered as it ‘occurred’ between observations. Either way, the sample is too small to yield any significant results.

Table 1 – Frequency table for resistance method variable

Change of primary resistance method (resis_meth)

Freq. Percent Cum.

0 1643 96.36 96.36

1 41 2.40 98.77

2 21 1.23 100.00

Total 1705 100.00

Source: NAVCO 2.0 dataset. Coding: 0=no change; 1=change from nonviolent to violent; 2= change from violent to nonviolent

Since the first variable does not take into account the changes in tactics and the third variable is rendered unusable due to the paucity of observations, the dependent variable selected for this

(27)

26 research is the second variable, “prim2”, which denotes the primary resistance method used in a campaign year. It is a dichotomous variable with 0 denoting a primarily nonviolent campaign and 1 representing a primarily violent campaign during that campaign year (Chenoweth & Stephan, 2011). Degrees of violence can be difficult to classify into these categories (Ackerman, Kruegler, Sharp, & Schelling, 1994) since campaigns may employ a combination of violent and nonviolent methods simultaneously, or can change methods of resistance over the course of a campaign. Nonetheless, by labeling the categories as primarily one or the other it allows scholars to classify the resistance method of a campaign by the majority of the tactics it employed over the course of the year. Nonviolent methods consist of resistance that does not directly threaten or harm the physical well-being of the opponent. Sharp’s (1973) classification of nonviolent campaigns is used, which includes almost 200 nonviolent resistance tactics such as boycotts, civil disobedience, protests, strikes, etc. The data on these nonviolent resistance methods was extracted by Chenoweth from notable sources, namely Ackerman and Karatnycky (2005), Carter, Clark, and Randle (2006), and Schock (2005). This data was cross-checked with multiple sources, including case studies, encyclopedias and experts in the field of nonviolent conflict, in order to evaluate the accuracy of the dataset and determine whether any observations were omitted.

Chenoweth obtains the data on violent campaigns predominantly from the 2004 version of Gleditsch’s Correlates of War (COW) database on intra-state wars, which requires a minimum of 1000 battle deaths to have occurred during the course of the conflict (Stephan & Chenoweth, 2008). Additionally, Clodfelter’s (2002) encyclopedia of armed conflict was used, as well as a list of major counterinsurgency operations composed by Sepp (2005) for data on conflicts after 2002. To make the dataset more comprehensive, numerous observations were added from Lyall and Wilson’s (2009) data on insurgencies, who list the COW database, the Uppsala Dataset on Armed Conflict, Fearon and Laitin’s (2003) dataset on civil wars and Clodfelter’s (2002) encyclopedia as their sources.

3.2.2 Basic independent variables

The variables in the NAVCO 2.0 dataset are categorized according to the labels defined in the conceptual model. Independent variables are characteristics or instruments of primarily either the state or the campaign, or when independent from these agents fall under the external category. The coding for the variables used in this research is shown in the Appendix.4

State

(28)

27 Progress

The NAVCO 2.0 dataset contains an ordinal variable which identifies if the campaign accomplished some or all of its stated overall political goals. This variable represents only progress towards overthrowing or receiving political concessions from the government, depending on the objective of the campaign. It also does not include “tactical or operational progress” which focuses on whether the campaign was able to achieve immediate tactical effectiveness, gather wider support or perform particular resistance actions. However the variable covers more actions than just concessions, and thus is termed as ‘progress’. The progress variable determines whether the campaign was effective in attaining political goals with respect to changes in the policy position and actions of the government in question. It is deemed the status quo if a state does not change its position at all. The variable is coded as “visible gains short of concessions” if the regime does not make formal or public concessions, but nevertheless alters its conduct to appease the opposition, such as allowing for more protest or political openness than was previously permitted. “Limited concessions” refers to verbal statements of conciliation or changes in the stated position of the government but a lack of any further action. Actual measures short of admission of surrender, such as expulsion of leading regime officials, policy changes or the initiation of negotiations with the opposition, represent “significant concessions”. Finally, the variable is termed as “complete success” if the government capitulates totally to the demands of the campaign. The coding is determined on the most noteworthy concession in a particular campaign year.

Repression

The repression variable in this dataset evaluates the most repressive period or activity conducted by the state in response to campaign actions. It observes repression from the perspective of the opposition campaign instead of the state. Concentrating on peak events, the variable measures the extent to which the regime uses the coercive apparatus of the state, such as the justice system, security agencies, and economic fines and taxes, in order to repress the opposition. The variable is categorized as “none” if the regime responds in a conciliatory manner to campaign activity, or does not respond at all. The variable is classified as “1” if there is mild repression present, which includes verbal or threatening actions short of physical action, economic fees or levies, or if the government expresses intent to engage in conflict. Moderate repression, coded as “2”, consists of physical or violent action by the state with the intent of coercion, harassment and imprisonment of campaign members, yet without an apparent intention to kill. Extreme repression is categorized as “3” – this is when the state employs physical action with intent to kill and violently silence opponents, torture and/or mass violence which could easily lead to deaths.

(29)

28 State Defections

The variable for state defections is a dichotomous, and determines if leaders associated with the state split with it to publicly announce their support for the campaign. If the variable is coded as 1 it implies that former state officials formally or tacitly support the campaign. Not included are officials that switch parties but remain within the existing ruling political framework. Neither are economic elites or other non-state actors, as these do not have a formal affiliation with the government. With “state officials” Chenoweth and Stephan (2011) refer primarily to the top non-military leadership of the state, such as cabinet ministers or notable political figures. This may also include prominent officials in diverse departments of the government that decide to cease working in their functions in order to support the campaign.

Discrimination

The discrimination variable focuses on the nature of state repression and how indiscriminate or precise a regime is in targeting the opposition campaign. Indiscriminate repression which does not discern between opposition members and general citizens, such as widespread state violence against the population or acts that threaten the use of force against all citizens, is coded as zero. The variable is denoted as one if state repression is discriminate and narrowly targets the opposition by for example closing down opposition organizations or arresting its leaders.

Campaign

Disunity

The disunity variable examines the presence of conflict within a campaign together with the nature and degree of conflict between groups, factions or leaders. It categorizes the nature of dissent and conflict within a campaign according to histories and narratives of the campaign and notable events that distinctly demonstrate the extent of conflict, for example intra-campaign attacks. The campaign may fall under numerous disunity categories throughout each year; however the peak code reached will apply for that particular year.

Diversity

The composition of large resistance movements can often be diverse in gender, age, ideology, ethnicity, religion, class, political party and geographical and even urban/rural distribution. Some forms of diversity can breed conditions for violence, whilst others may propagate nonviolent

Sticks and stones : the causes of collective violence in protest movements