Gender Equality and Nonviolent Political Campaigns

Gender Equality and Nonviolent Political Campaigns Laura Huber Emory University March 9, 2016 Prepared for the 57th ISA Annual Convention 2016, Atlanta Georgia Please do not cite without the author s permission. Abstract Nonviolence has been increasingly viewed as a successful tactic to call for political change. However, the state- and campaign-level characteristics that promote nonviolence are still understudied. This article examines one particular characteristic that promotes nonviolence: gender equality. Gender equality increases the strategic value of nonviolence by decreasing militarism and increasing norms of tolerance and respect, both of which lead the government and society to be positively responsive to nonviolence and decrease the utility of violence. To analyze the relationship between gender equality and nonviolence, I examine the influence of state-level women s rights and campaign-level gender diversity on the choice of nonviolent tactics among contentious political campaigns. Further, I examine how state-level indicators of gender equality influence the success of nonviolent campaigns. Using the NAVCO 2.0 dataset and OLS regression, I demonstrate that state-level women s rights and campaign-level gender diversity increase the probability that a contentious political campaign will use nonviolent tactics rather than violent tactics. However, the relationship between state-level gender equality and the success of nonviolent political campaigns is less clear. Overall, this study increases our understanding of the structural characteristics that promote nonviolence. Ph.D Student at Emory University. lkhuber@emory.edu 1555 Dickey Drive Atlanta, Georgia 30322

Introduction Since the end of the Cold War, nonviolence has been increasingly viewed as a successful tactic to challenge the state. From the Color Revolutions in Eastern Europe to the nonviolent demonstrations of the Arab Spring, nonviolent movements have sparked major political changes around the globe. However, violent conflict persists, often coinciding with nonviolent conflict. Overall, nonviolence remains a relatively understudied phenomenon, especially in comparison with violent conflict. This study seeks to expand our understanding of the interaction between nonviolence and campaign- and state-level characteristics by specifically examining the effect of gender equality on the behavior of domestic contentious political campaigns. Gender equality has consistently been connected with decreased levels of violence, both anecdotally and empirically. Gender equality has been shown to decrease the occurrence and intensity of inter- and intra-state violence, discourage human rights abuses, and promote democratic and economic development (Caprioli 2005, Melander 2005a, Caprioli and Boyer 2001, Melander 2005b, Caprioli 2003 Dumas 2010, Hudson et al. 2012). However, an outstanding question remains. What is the relationship between gender equality and nonviolent protest. This study begins to address this question by investigating how gender equality affects the choice and success of violent and nonviolent contentious tactics. First, I examine whether state-level gender equality and campaign-level gender diversity encourage nonviolent political behavior among contentious political campaigns. Gender equality decreases structural violence and militarism, promotes interpersonal norms of respect and tolerance, and increases the potential recruitment pool of campaigns, which may lead to more opportunity for nonviolent mobilization to address perceived injustices. Therefore, state and campaign gendered dynamics may increase the perceived utility of nonviolent tactics relative to violent tactics. Then, I examine how state-level gender equality interacts with campaign-level dynamics. Specifically, the potential positive relationship between campaignlevel gender diversity and state-level gender equality are mutually reinforcing. Lastly, I move 2

beyond the organization s choice of nonviolent tactics to examine whether these tactics are successful. Gender equitable states are more likely to have governments and citizens that are more favorable to nonviolent protest and less likely to repress nonviolent movements. Therefore, nonviolent campaigns in gender equitable states may be more successful than nonviolent campaigns in inequitable states. Using OLS regression and the Nonviolent and Violent Campaigns and Outcomes(NAVCO) Data Project (Chenoweth and Lewis 2013), I explore the relationship between campaign and state gender characteristics and the campaign s tactical choices and success. The findings demonstrate that state-level gender equality and campaign gender diversity promote nonviolent behavior. Surprisingly, the findings also indicate that state-level gender equality may decrease the success of nonviolent tactics. These findings present an interesting theoretical and empirical puzzle that underscore the need for further investigation into how nonviolent behavior differs from violent behavior. To the author s knowledge, this is the first crossnational quantitative study to examine how state-level gender equality and campaign gender diversity influence contentious campaigns nonviolent behavior. The paper proceeds as follows. First I briefly discuss the use of nonviolent contention as a strategic choice. Next, I propose that state and campaign gender equality should promote nonviolence, discourage violence, and increase the success of nonviolent campaigns. Finally, I present a research design and discuss the results before concluding. Nonviolent Political Contention as a Strategic Choice While the study of violent conflict has often taken precedence over the study of nonviolence, scholarship on nonviolenct conflict has increased in recent years (Chenoweth and Stephan 2011, Lawrence 2010, Pearlman 2011). Importantly, scholars have emphasized that nonviolence is a strategic political tactic itself, rather than an essentialist group characteristic or a step towards violent escalation (Chenoweth and Stephan 2011, Asal et al. 2013). There is a dynamic field of work on nonviolent civil resistance, which highlights the 3

strategic logic of nonviolent political behavior. Sharp (1973) and Helvey (2004) examine the strategy of civil resistance, analyzing the different types and tactics of nonviolence and their effects on the adversary. Moreover, Ackerman and DuVall (2000) examine how nonviolent movements successfully challenge despotic regimes, strengthen democratic norms, and protect human rights through a number of case studies. Schock (2004) argues that the use of nonviolence is a strategic choice and that outcomes of nonviolent movements can be explained by a combination of political opportunities and strategic choices. Moreover, Chenoweth and Stephan (2011) conduct the first cross-national quantitative analysis of nonviolent campaign success. Using the NAVCO 1.1 dataset, they demonstrate that nonviolent campaigns are more successful on average that violent campaigns. They argue that this success is due to the larger size and diversity of the participants of nonviolent campaigns. Similarly, Chenoweth and Lewis (2013) argue that nonviolent and violent campaigns emerge from different structural factors, with nonviolent campaigns more commonly mobilizing in difficult contexts. Asal et al. (2013) examine how group characteristics influence the choice of violent, nonviolent, and mixed (nonviolent and violent) tactics. They claim that organizations in democratic states and those with positive gender ideologies are more likely to use nonviolence and that organizations that are repressed, have diaspora support, and hold a leftist ideology are more likely to use violent methods. Finally, Chenoweth and Ulfelter (2015) test the predictive power of the grievance, modernization, resource mobilization, and political opportunities models of violent conflict to explain nonviolent campaign onset and find moderate, although uneven, support for each model, but conclude that no single model can accurately predict nonviolent campaign onset. 1 Importantly, literature on nonviolent political movements has consistently demonstrated that nonviolence is a strategic choice made by political movements in response to stateand campaign-level structural and ideological factors. This contrasts previous logic that 1 Specifically, they find that nonviolent campaign onset can be partially explained by poverty levels, the leader s tenure in office, urbanization, communication technology, contagion pressures, youth bulges, social unrest, civil liberties, election years, the post-cold War period, and being a signatory to the Optional First Protocol of the International Covenant on Civil and Political Rights. 4

nonviolent actions were either the result of moral restrictions on group behavior or occupied the low-end of the continuum of political contentious behavior, acting as a precursor to violence. Moreover, these studies highlight the success of nonviolent campaigns in a variety of contexts against differing opponents. Moving forward, this study specifically examines nonviolent campaigns that represent contentious political action. Contentious politics or contentious political behavior refers to political action in which the actor or organization leaves the realm of traditional political channels to challenge the state and gain some concession from it (McAdam et al. 2007, Tilly 2008, Asal et al. 2013). Contentious political behavior refers to both violent actions, such as bombings and armed attacks, and large-scale disruptive nonviolent actions, such as demonstrations and protests. Contentious tactics refer to the contentious entity s choice of using violence or nonviolence. Therefore, nonviolent campaigns are a form of political contention in which a contentious group of individuals chooses to use massive and disruptive nonviolence and behavior (as opposed to violence) to coerce the state into granting concessions. A contentious campaign refers to the movement or organization that represents or carries out the violent or nonviolent actions against the state. Gender Equality and Diversity: Facilitating Nonviolence Political contentious campaigns seek to coerce the government to grant concessions through public, often disruptive, action. Contention is a costly activity, both in terms of the resources needed to organize and support the campaign and in terms of the opportunity cost or possible threat of punishment for challenging the government. Therefore, the leaders of well-organized political campaigns, as strategic actors, will choose the type of contentious tactic that they believe will be most effective against the targeted government at the lowest cost. The perceived utility of using violence relative to nonviolence can be affected by several factors, such as campaign and government resources or ideology. This study examines one factor that may alter perceptions of the utility of violence and nonviolence, gender dynamics. 5

State-Level Gender Equality and Nonviolent Campaigns Scholars and societies have long theorized that women encourage peace and deter violence in their communities (Goldstein 2001). Generally, Women, Peace, and Security (WPS) literature argues that as gender equality increases, societies become more peaceful and less violent. Caprioli (2005) and Melander (2005a) demonstrate that intrastate violence is less likely and less intense in gender equitable states. Similarly, Caprioli and Boyer (2001) find the same relationship between gender equality and interstate violence. Melander (2005b) also finds that high levels of state gender equality predicts lower levels of human rights violations. Caprioli (2003) demonstrates that domestic gender equality decreases a state s willingness to use violent force. High levels of gender equality have also been linked to decreased incidences of domestic terrorism (Dumas 2010). Further, Hudson et al. (2012) demonstrate that gender equality not only improves economic planning and reduces corruption, but that it also is an important piece in the groundwork for democracy. Importantly, gender equality is increasingly viewed as a separate, independent driver of peace, rather than a by-product of other peace-promoting factors, such as democracy, wealth, or culture (Melander 2005b). Generally, scholars argue that the beneficial impact of gender equality lies in its transformative effect on interpersonal relations and societial norms. Melander (2005a) claims that gender equality promotes mutual respect between individuals regardless of gender and that these positive interpersonal relationships spill over to larger societal patterns of tolerance, respect, and peacefulness. Caprioli (2005) also claims that gender inequality is a form of structural violence, which results in more instability and societal violence. 2 However, while gender equality s negative influence on violence has been well-established, less is known about its affect on nonviolent behavior. Women and gender equality have 2 Additionally, essentialist claims argue that gender equality results in less violence because women are inherently peaceful and less violent than men whereas men are aggressive and militaristic by their nature. Therefore, these theories argue that when women are empowered in society and share equal rights with men, they can exert more of their natural peaceful influence. However, essentialist claims have largely been rejected in favor of more constructivist claims in light of overwhelming evidence that women are not inherently peaceful (and men are not inherently violent). For examples of evidence against the essentialist view of women and men s attitudes towards violence, see Goldstein 2001, Bloom 2011, and Cohen 2013. 6

long been theorized to support nonviolence. Goldstein (2001) discusses women s traditional association with peace and the feminization of nonviolent movements. Similarly, Enloe(2007) argues that patriarchal values promote militarism and encourage violence, while feminist perspectives tend to favor nonviolent resistance. Whether due to social conditioning or to the dismantling of patriarchal power structures that encourage competition and conflict, gender equality is widely believed to promote nonviolent behavior. Specifically, I argue that patriarchy and gender inequality is a form of structural violence that establishes violence as the prima facie way of achieving order and stability in society or to achieve political and social success. Patriarchy is a rigid societal hierarchy that discounts half of society for their physical characteristics while privileging others. This societal structure establishes norms of dominance and subjugation. It also requires constant maintenance with violence, both physical and mental, overt and subtle, to prevent women (and subjugated men) from asserting themselves against the patriarchal order and hegemonic masculinity (Caprioli 2005). Further, values of militarism and aggression become accepted and celebrated as the necessary characteristics for success (Enloe 2007). As a result, violence becomes the normal way of enforcing order and solving interpersonal, intergroup, and interstate disputes in gender inequitable societies. In contrast, gender equitable societies face a less rigid patriarchal order, that discounts the hypermasculine values required to sustain high levels of gender inequality, such as violence, militarism, and aggression, and instead upholds more traditionally feminine values, such as patience, tolerance, andnonviolent disputeresolution. 3 Further, ingenderequitable societies, violence is no longer, or less frequently, used as the main vehicle to control social organization and interpersonal relations (Caprioli 2005). Without high levels of gender inequality acting as a constant form of structural violence, which installs violence as a societal norm, violence may simply no longer act as the normal way to solve disputes. Further, since gender equitable states are less militaristic and aggressive, violence may be less commonplace. As a result, 3 It should be noted that I am not claiming that women inherently possess these characteristics, causing them to be feminine. Instead, these characteristics are generally those associated with women due to the social construction of gender roles and therefore, discounted, by patriarchal societies. 7

when disputes arise, the parties to the conflict should be less inclined to choose violence, since it is discredited as a proper tactic, and should be more likely to choose nonviolent dispute resolution methods, the more commonly accepted mode of behavior and resolution in gender equitable states. Therefore, gender equality shifts perceptions of the utility of violence and nonviolence as a contentious tactic. This effect occurs at two levels. On one hand, the non-state political actors of the campaign may themselves be more supportive of nonviolent behavior. If gender equality affects individual beliefs and attitudes by decreasing aggression and militarism (Melander 2005a, Caprioli 2005), then the leaders and members of political campaigns in this environment may view nonviolent tactics as a more appropriate and tactically successful choice than violent behavior. On the other hand, the target of the campaign, which in this study is the government of the state, may also be more supportive of and responsive to nonviolent tactics if gender equality has led it to be less militaristic and patriarchal. Therefore, nonviolent tactics may not only align more with the norms and values of a gender equitable society and possibly the campaign participants, but it may also be a more appropriate strategy for the campaign to employ against the state. Therefore, gender equality should promote nonviolent tactics over violent tactics because it increases the strategic value of nonviolence. State-level gender equality can be measured in various ways, each of which may have differing implications for political action. This study specifically examines women s aggregate economic, political, and social rights as a proxy for gender equality. Hypothesis 1: Political campaigns should be more likely to choose nonviolent tactics than violent tactics in states with high levels of women s rights. Campaign Gender Diversity and Nonviolent Campaigns The peaceful influence of gender equality, of expressed and observed through gender diversity, is also observed at the group level. For example, groups with larger proportions of women 8

are more likely to be nonviolent and less likely to be violent than male-dominated groups (Warren & Caddy 1994, Costain 2000, Beckwith 2002, Zimmerman et al 2009). Further, Asal et al. (2013) demonstrate that ethno-political groups in the Middle East and North Africa which have positive gender ideologies, or support the inclusion of women in public life, are more likely to use nonviolence than groups with a positive gender ideology. The exact relationship between group gender diversity and nonviolence is somewhat unclear. Some argue that it arises because women are more favorable towards nonviolence and therefore, when women s representation within the group increases, their opinions should craft group behavior, assuming that their opinions are taken into account by group decisionmakers. 4 For example, Gilligan (1997) argues that women s moral judgment evaluates nonviolence as the most adequate resolution of conflict. Moreover Ruddick (1997) argues that women s maternal role and its contingent emphasis on caregiving and protection leads them to be more supportive of nonviolence. These arguments tend to point to the near absence of violence among women s movements as examples of women s encouragement of nonviolence in political behavior. Other perspectives imply that women s participation in organizations leads to less violent behavior not directly due to women s inherent desire for nonviolence, but because it leads to a diversity of opinions and perspectives within a group, which decreases the likelihood of radicalization (Chenoweth and Stephan 2011). Further, it potentially signals that the campaign itself holds gender equitable values that may indicate that it is less likely to condone violence. Whether campaign gender diversity leads to women s less aggressive opinions to shape tactic adoption, signals campaign-level gender equality, or results in more campaign heterogeneity that makes the adoption of violent tactics less feasible, gender diversity should increase the probability that a political campaign adopts nonviolent methods. Hypothesis 2: Gender diverse campaigns should be more likely to use nonviolent tactics than campaigns that are not gender diverse. 4 Manycritics ofthe justaddwomenandstir approacharguethatsimplyaddingwomentoanorganization does not ensure that their opinions will be taken into account when decisions are made or that the organization will begin to reflect more gender equitable values. For examples of this argument, see Dharmapuri (2007). 9

However, gender diverse does not necessarily ensure that the campaign holds a gender equitable ideology or that it rejects violence. For example, women often participate in terrorist groups as agents, suicide bombers, informants, and supporters (Bloom 2011). Moreover, Thomas and Bond (2015) demonstrate that violent political organizations commonly include women, often in response to strategic challenges that may require the use of women s unique skills or to a lack of male recruits. Therefore, campaign gender diversity may not always lead to nonviolence. However, campaign gender composition and state-level gender equality may interact to increase the probability of nonviolence. First, whencampaigns are diverseand operate in astate with high levels of gender equality, this may be a better indication that the campaign itself holds relatively positive gender norms. While an imperfect proxy for campaign gender ideology, considering both campaign composition and societal gender equality may allow us to approximate campaign-level gender equality more perfectly than simply examining campaign diversity. Second, campaign gender diversity may enhance the likelihood that the organization will select nonviolence as a way to gain the support of the citizens and the acquiescence or acceptance by the state, under the assumption that more gender equitable communities are more responsive to nonviolent methods than violent methods due to the decrease in militaristic, aggressive, and violent values in society. Hypothesis 3: Gender diverse campaigns should be more likely to use nonviolent methods in states with high levels of women s rights. Gender Equality and The Success of Nonviolent Campaigns Finally, if nonviolence is a strategic choice in gender equitable states, then campaign decisionmakers must expect that nonviolent methods will be more successful and less costly than violent methods in this context. As explained above, gender equitable states should be more supportive of nonviolence than violence. Therefore, governments in gender equitable states may be more willing to negotiate with organizations using nonviolence than violence 10

since these governments should theoretically reject the efficacy of violence as a method of conflict resolution. This counters some common arguments that gender equitable states, which are usually less militaristic, should be weaker and more likely to fold in response to violent challenges. For example, Dube and Harris (n.d.) propose that unmarried queens who reigned between the fifteenth and twentieth centuries were the victims of aggression relatively more than kings or married queens because they were perceived to be weaker and more doveish. Similarly, it is believed that Saddam Hussein may have not heeded the warnings of the U.S Ambassador to Iraq of retaliation for aggression against Kuwait because she was a woman. However, despite the commonality of this logic, very little evidence supports the claim that gender equitable states are in fact more likely to be cowed into submission by violent acts or that female leaders are less aggressive than men. For example, Indira Gandhi was reportedly eager to intervene in the Pakistani civil war, even with the discouragement of her own generals and the United States (Gupte 1992). Similarly, other female heads of state, such as Margaret Thatcher, were known by their male counterparts to be steadfast and ironwilled. Further, Koch and Fuller (2011) find that female executives are more likely to act aggressively in defense spending than male leaders. Therefore, the actual behavior of female leaders does not appear to confirm the dove-ish stereotype. However, it is important to note that female leadership does not equate to gender equality. Studies have consistently shown that while societal indicators of gender equality decrease violent behavior by the state, having a female leader as the head of state or chief executive does not decrease violent behavior, and at times may increase aggressive behavior by the state (Koch and Fuller 2011). I claim that nonviolent campaigns may be more successful in gender equitable states because the government is more likely to make concessions to campaigns using accepted nonviolent methods, which are viewed as the condoned way to solve disputes, rather than condemned violent methods. In contrast, in less gender equitable states, where militaristic values and violent conflict resolution are more accepted, states may be more likely to respect the challenge of violent campaigns and reject the weak methods of nonviolence. 11

Moreover, campaigns using nonviolent methods in gender equitable states are also more likely to gain wider support from the population, which may translate into greater levels of participation in the campaign and more pressure on the government to make concessions. On one hand, and similar to the argument above, in gender equitable states, citizens may be more supportive of nonviolent methods than violent methods due to societal norms. Therefore, they may be likely to view a political campaign that uses nonviolent methods more favorably than a violent campaign. This indirect popular support may lead the government to feel more pressured to negotiate with the campaign and make concessions. Similarly, societal gender equality may lead to more direct support for campaigns using nonviolence. For example, if more people support the campaign, it may receive more resources and participants. Moreover, gender equality indicates that women are more active in society, and therefore, may be more likely to join the campaign, making it larger and more formidable against the government. Therefore, contentious campaigns should not only be more likely to use nonviolent tactics in gender equitable states, but nonviolent campaigns should be more successful at attaining their goals than violent campaigns. Hypothesis 4: Nonviolent campaigns should be more likely to succeed than violent campaigns in states with high levels of women s rights. Research Design In order to test the influence of state and campaign gender characteristics on nonviolent behavior, I primarily use the Nonviolent and Violent Campaigns and Outcomes (2.0) Dataset (NAVCO) and the CIRI human rights datset. NAVCO is a multi-level data project that examines violent and nonviolent political campaigns across the globe. Specifically, I am using NAVCO 2.0, which contains data on annual contentious campaign behavior from 1945-2006 for a total of 1,174 campaign years from 250 campaigns (Chenoweth and Lewis 2013). 12

Each campaign had over 1,000 members at its peak and a coherent organizational structure (Chenoweth and Lewis 2013). Therefore, they represent mass, large-scale contentious behavior, rather than lower-level nonviolent protest. Further, I restrict the sample to only domestic political contentious campaigns. Therefore, my final sample includes 1,111 violent and nonviolent political campaign years from 248 campaigns in 95 countries. 5 NAVCO 2.0 represents the most comprehensive sample of large-scale contentious behavior. However, due to reporting bias, it is likely that nonviolent campaigns and unsuccessful campaigns are underrepresented in the sample. Therefore, these results represent an exploratory examination of the relationship between gender dynamics and nonviolence, rather than a definitive test of causality. Throughout the analysis, I use linear regression with state fixed effects to control for unobserved heterogeneity among states that may be influencing contentious campaign tactics and success. Although nonlinear models are a common method to estimate dichotomous dependent variables, as used in this analysis, they do not correct for unobserved heterogeneity among units in panel data and can suffer from the incidental parameters problem with fixed effects (Neyman and Scott 1948, Carmen and Trivedi 2005). Linear regression recovers the marginal effects of nonlinear models, allows the use of fixed effects, and facilitates interpretation (Angrist and Pischke 2009). Since I believe unobserved heterogeneity to be a major concern, the benefit of using linear regression with fixed effects overcomes concerns of using a linear model with a nonlinear dependent variable. Therefore, I use ordinary least squared regression throughout the main analysis with state fixed effects to account for unobserved heterogeneity at the state level that may influence tactical choices. However, all analyses are also run with logistic regression, which can be seen in the appendix. 5 The major campaigns years removed from the dataset were campaigns against imperialist or colonial foreign powers, many of which would have been dropped from the analysis regardless due to missing data. 13

Gender Equality and Nonviolence To test my first hypothesis, that state-level gender equality increases the probability of nonviolence, I use as the dependent variable a dichotomous indicator from the NAVCO 2.0 dataset of the primary tactic, violent or nonviolent, used by the campaign duringthe year. It is coded as 1 if the campaign primarily uses nonviolent tactics and 0 if the campaign uses primarily violent tactics. Nonviolent tactics were identified as those that that do not directly threaten or harm the physical well-being of the opponent, whereas violent tactics aim to physically harm or threaten the opponent (Chenoweth and Lewis 2013). Nonviolent campaigns include tactics such as sit-ins, protests, boycotts, civil disobedience, and strikes. While campaigns often use a mixture of violent and nonviolent tactics, NAVCO 2.0 focuses on the primary tactic of a campaign. For example, if a campaign relies primarily on nonviolent methods as opposed to violent or armed tactics, the campaign can be characterized as nonviolent. However, a coding of nonviolent does not mean that the campaign never uses violent methods, only that nonviolent tactics are more commonly used than violent methods. Campaign tactics can vary from year to year. The main independent variable, Fem Rights is an ordinal, composite measure of women s rights in the state-year. To create this measure, I used three indicators of women s rights from the CIRI Human Rights Dataset (Cingranelli et al. 2013). First, women s political rights are captured in WOPOL, an ordinal measure ranging from 0 (no political rights guaranteed by law) to 3 (women s political rights guaranteed in practice and by law). CIRI defines women s political rights as the right to vote, run for political office, hold elected and appointed government positions, join political parties, and the right to petition government officials. Second, WECON measures women s economic rights in the country, including the right to equal pay for equal work, free choice of profession, equality in hiring and promotion practices, job security, non-discrimination by employers, freedom from sexual harassment, and the ability to work in various dangerous or traditionally male professions. WECON also ranges from 0 to 3 with 0 indicating that women have no economic rights and that 14

systematic sex-based discrimination may be built into the law and 3 indicating that women s economic rights are guaranteed by law and in practice. Finally, WOSOC is a four level ordinal variable from 0 to 3 that measures women s social rights, such as right to inheritance, marriage, international travel, citizenship, divorce, property, education, public participation, and freedom from female genital mutilation and forced sterilization. 6 Using these three indicators of women s rights, I created an additive index of general women s rights in the state, which ranged from 0 (if the state scored 0 in each component of women s rights) to 9 (if the state scored 3 on each component). This variable is lagged by one year to address concerns of endogeneity. It remains lagged throughout the remainder of the models. Finally, I include several control variables. I control for the regime type of the state by using the raw Polity II measure, Polity2, and the squared Polity II term, Polity 2, to account for the nonmonotonic effect of democracy on violence and political action (Marshall et al. 2013). When democracy is very low or very high, there may may be limited ability or will to use violence or nonviolence, but in semi-democracies there is both will and opportunity, which results in higher levels of contentious action (Muller and Weede 1990, Hegre et al. 2001, Elbadawi and Sambanis 2002). Furthermore, democratic states have often been linked with gender equality. Polity2 and Polity 2 are both lagged by one year to account for autocorrelation. Second, I control for the gross domestic product (GDP) of the state since wealthier states are less likely to experience violent conflict and more likely to have higher levels of gender equality (Hudson et al. 2013). GDP is obtained from the World Bank, log transformed, and lagged by one year to account for autocorrelation. Moreover, I control for the population size of the state since larger populations may be more likely to use nonviolence (Chenoweth and Stephan 2013) and may influence the level of women s rights in the country since larger populations may indicate greater demand for women s rights. This control is taken from the World Bank and lagged by one year. I include a control for whether the state experienced an inter- or intra-state war, according to the Correlates of War dataset (Sarkees and Wayman 2010). Participation in war is likely 6 This variable was retired in 2005. Therefore, observations after 2005 are eliminated from the sample. 15

to increase the instability of the state, which may lead to more violence. Further, conflict has been linked to changes in women s rights. On one hand, war may increase economic opportunity for women if they are recruited into the workforce to replace male soldiers or into newly developed industries to support the war (Goldstein 2001). However, more commonly, war has a negative influence of women s rights since women often suffer from economic and social displacement. This variable, War is also lagged by one year. I also include a dummy variable for the Cold War era since nonviolent campaigns became more common after the Cold War and women s rights also tended to increase after the Cold War. Finally, to account for autocorrelation within the dependent variable, I include two control variables, Time Since Violence and Time Since Violence 2. Since the use of violence is correlated between years within a campaign, the use of violence (or nonviolence) by a campaign one year, strongly predicts the use of violence (or nonviolence) the next year. I use the approach by Carter and Signorino (2010) who use a cubic polynomial, which captures the hazard rate of the use of violence. More specifically, it captures any hazard shape that is recovered by commonly estimated parametric duration models such as a Weibull or loglogistic model. This approach avoids problems associated with time variable dummies and with interpreting splines. Gender Diversity and Nonviolence To test my second hypothesis, that gender diverse campaigns should be more likely to use nonviolent methods, I use the same dependent variable, Nonviolent Tactics, and linear regression. My independent variable is a dichotomous measure of whether the campaign, in the campaign year, was diverse in the gender of its members. The variable, gathered from the NAVCO 2.0 dataset, Gender Diverse, is coded as 1 if the campaign was gender diverse and 0 if it was not diverse. Diversity is not strictly defined with a specific threshold, but a campaign is coded as being gender diverse when there is evidence that membership spans the socio-demographic category (Chenoweth and Lewis 2013). My theory expects that gender 16

diverse campaigns should be more likely to adopt nonviolent methods and therefore, there should be a positive relationship between Gender Diverse and Nonviolent Tactics. Furthermore, I include several control variables. First, I consider campaign-level characteristics that may influence both whether the campaign uses nonviolent tactics and whether the campaign is gender diverse. Camp Size is an ordinal measure of the estimated size of the campaign in the year. The variables ranges from 0, meaning the campaign had hundreds to thousands of participants, to 3, meaning the campaign had over one million participants. 7 As the campaign size increases, the campaign becomes more likely to use nonviolent methods and less likely to use violent methods since the campaign is less likely to become radical and is more likely to receive internal and external support and security force defections, reducing the need for violence (Chenoweth and Stephan 2013). Moreover, as the size increases, women are more likely to be included, increasing the gender diversity of the campaign (Thomas and Bond 2015). Moreover, I include a control for whether the campaign was repressed by the state in the campaign-year. Repression is an ordinal measure of state repressive responses to campaign activity which ranges from 0, no repression to 3, extreme repression. 8 Asal et al. (2013) demonstrate that repressed campaigns are more likely to use violent tactics. Further, repression may decrease the gender diversity of the campaign by preventing or dissuading women from participating. Therefore, repression should have a negative relationship with both nonviolence and gender diversity. I also include a control for whether the campaign received support from an external state, Camp Support. External state support affects tactical choice by providing more resources to the campaign to use in either violent or nonviolent tactics. Moreover, since this variable is restricted to public state support, it is likely to be correlated with nonviolent tactics since international norms favor nonviolent political behavior. Moreover, external support may be viewed as increasing the legitimacy of the 7 According to the NAVCO 2.0 codebook, 0=small (hundreds to thousands) 1=medium (tens of thousands) 2=large (above one hundred thousand) 3=extremely large (above one million) 8 AccordingtotheNAVCO2.0 dataset, repression is codedin thefollowing way. 0=none; 1=mild repression; 2=moderate repression; 3=extreme repression. 17

campaign, which may lead to greater and more diverse participation. Finally, I include several state structural controls that may influence the choice of tactics andthegenderdiversityofcampaigns, includingpolity2, Polity 2, War, Cold War, Population, and GDP, which are all lagged by one year, except Cold War. textitgdp and Population are log-transformed. Finally, to account for autocorrelation in the DV, I include controls for Time Since Violence and Time Since Violence 2. As before, I include state fixed effects to account for any unobserved characteristics of the state in which the campaign is operating (and targeting since this analysis only looks at domestic political campaigns) that may affect the adoption of nonviolent tactics. To test the relative strength of the two mechanisms in comparison with each other, I also run a model in which both Fem Rights and Gender Diversity are included as the main independent variables. I include controls which influence both IVs, including Polity2, Polity 2, GDP, Population, War, Cold War, Time Since Violence, and Time Since Violence 2, and state-fixed effects. Finally, to test the interactive effect of campaign diversity and state-level gender equality, Iuseaninteraction modelwhereiinteract mymeasureof women srights withtheindicator of campaigngenderdiversity. Asbefore,IincludecontrolsforPolity2, Polity 2,GDP,Population, War, Cold War, Time Since Violence, and Time Since Violence 2. State fixed effects are included to account for heterogeneity that affect tactical choices. Gender Equality and the Success of Nonviolent Campaigns Finally, to test my last hypothesis that gender equality should increase the success of nonviolent campaigns, I use a dichotomous measure of campaign success in the campaign-year. A campaign-year is coded as 0 if in that year the state does not change its position toward the campaign, does not grant any concessions to it in that year, grants visible gains short of concessions, or gave significant concessions, but did not offer complete success. A coding of 1 indicates that the campaign achieved complete success in the year. This variable was 18

coded according to the NAVCO 2.0 dataset. My independent variables are Fem Rights and Nonviolent Tactics. Since my theory predicts that nonviolent tactics may be dependent upon the level of gender equality in a state-year, I use an interaction model. I use OLS regression with state fixed effects. Finally, I include control variables. Several of these, GDP, Cold War, Population, Polity2, and Polity 2 have already been described. However, I include several new variables. First, CINC is a composite measure of the state s national material capabilities, which includes total population, urban population, iron and steel production, energy consumption, military personnel, and military expenditure (Singer et al. 1972). States with greater capabilities may be able to effectively counter political campaigns, decreasing the probability of success. Moreover, national material capabilities may also be correlated with the level of gender equality, especially as indicators of military strength may point to militaristic, patriarchal cultures. CINC is lagged by one year to account for autocorrelation. Second, Sanctions is a dichotomous indicator of whether the international community imposed sanctions on the target regime in response to the state s repression of the campaign. Sanctions may make a nonviolent campaign more likely to win and may increase the probability that a campaign chooses nonviolent tactics in response to international norms and pressure (Chenoweth and Stephan 2013). It also may decrease the level of gender equality in the country since sanctions often lead to worsened economic conditions that may have negative effects on gender equality. Finally, I include a one year lagged dependent variable to account for autocorrelation within the dependent variable. Summary statistics for the main variables can be seen in Table 1 below. 19

Table 1: Summary Statistics Statistic N Mean St. Dev. Min Max Nonviolent Tactics 1,111 0.240 0.427 0 1 Fem Rights 809 3.738 1.373 0 8 Gender Diverse 1,032 0.822 0.383 0 1 Polity2 1,006 0.719 6.256 9 10 Polity 2 1,006 39.617 25.854 0 100 GDP 573 24.351 2.115 18.700 27.779 Population 638 17.285 1.585 12.601 20.858 War 1,111 0.416 0.493 0 1 Cold War 1,111 0.410 0.492 0 1 Time Since Violence 1,111 0.914 2.916 0 26 Time Since Violence 2 1,111 9.331 51.025 0 676 Repression 1,095 2.644 0.859 0 3 Camp Size 1,009 1.626 1.016 0 5 Camp Support 1,013 0.533 0.499 0 1 CINC 1,042 0.012 0.019 0.00002 0.109 Sanctions 1,089 0.241 0.428 0 1 Success 966 0.034 0.182 0 1 Results and Discussion Gender equality has long been theorized to have a positive relationship with nonviolence, however, relatively little is known about how state-level gender equality and campaign-level gender dynamics influence contentious tactical choices. The results show that state-level gender equality and campaign-level diversity both promote the use of nonviolent tactics. Models 1 through 4 test the the influence of state-level and campaign-level gender dynamics on the probability that a contentious campaign uses nonviolent tactic. Model 5 tests the influence of state-level gender equality on the probability that the campaign is successful. The results of all of the models can be seen in Table 2. Model 1 demonstrates the relationship between women s rights and the likelihood that an organization chooses nonviolent tactics. As expected, we find a significant positive relationship, which indicates that in gender equitable states political campaigns are more likely to choose nonviolent than violent tactics. As women s rights increase by one unit on the 9 20

point scale, campaigns become 4 percent more likely to use nonviolence instead of violence. This relationship is statistically significant at the 5% level. This supports my theory that state-level gender equality influences campaign behavior by decreasing the utility of violence and increasing the utility of nonviolent tactics. Model 2 presents the results of the influence of campaign gender diversity on the choice of nonviolent tactics. In compliance with my expectation that gender diversity has a positive influence on nonviolent tactical adoption, Model 2 demonstrates that campaign-level gender diversity has a positive and statistically significant relationship with nonviolent behavior among contentious political campaigns. Further, gender diversity has a large substantive effect, increasing the probability that a campaign will use nonviolent tactics by 34 percent. This supports my hypothesis and previous theories that gender diversity within an organization should decrease aggression and militarism and promote nonviolent behavior (Enloe 2007). However, this finding does not deny that gender diverse political campaigns may use violent tactics, as can be witnessed by the female fighers of the IRA in the 1970s or the Tamil Tigers, but instead indicates that gender diversity will tend to make nonviolence more likely. Moreover, the pacifying effect of gender diversity appears to act separately from the size of the movement. In other words, gender diverse campaigns do not simply happen to also be larger campaigns, with the large size leading to nonviolent behavior (Chenoweth and Stephan 2013), but that gender diversity is a unique mechanism promoting nonviolence. Model 3 presents the full model which includes both women s rights at the state-level and the gender diversity of the campaign to facilitate comparison between these two mechanisms. As can be seen, both gendered mechanisms have an even stronger effect when considered simultaneously. As state-level women s rights increase, political campaigns become 7 percent more likely to use nonviolent tactics. Further, gender diverse campaigns are 43 percent more likely to choose nonviolent tactics relative to non-diverse campaigns. Both of these findings are highly statistically significant at the 1% level. These results further underscore the highly influential role of gendered dynamics on campaign tactical choices. Further, they assuage fears that the two mechanisms are proxies for one another. 21

Table 2: OLS Regression Results 22 Dependent variable: Nonviolent Tactics Success (1) (2) (3) (4) (5) Fem Rights 0.04 (0.02) 0.07 (0.02) 0.04 (0.03) 0.03 (0.02) Gender Diverse 0.34 (0.05) 0.43 (0.05) 0.31 (0.14) Nonviolent Tactics 0.58 (0.14) Polity2 0.003 (0.004) 0.01 (0.004) 0.01 (0.003) 0.01 (0.004) 0.003 (0.003) Polity 2 0.002 (0.001) 0.001 (0.001) 0.001 (0.001) 0.001 (0.001) 0.002 (0.001) GDP 0.06 (0.04) 0.004 (0.04) 0.04 (0.04) 0.04 (0.04) 0.02 (0.04) Population 0.14 (0.08) 0.08 (0.07) 0.14 (0.07) 0.14 (0.07) 0.04 (0.07) War 0.03 (0.03) 0.03 (0.03) 0.02 (0.03) 0.02 (0.03) 0.04 (0.03) Cold War 0.10 (0.04) 0.04 (0.04) 0.05 (0.04) 0.05 (0.04) 0.07 (0.04) Repression 0.07 (0.01) Camp Size 0.07 (0.02) Camp Support 0.02 (0.03) Time Since Violence 0.21 (0.02) 0.22 (0.02) 0.20 (0.02) 0.20 (0.02) Time Since Violence 2 0.01 (0.002) 0.02 (0.003) 0.01 (0.001) 0.01 (0.001) Gender Diverse:Fem Rights 0.03 (0.03) CINC 2.09 (3.13) Sanctions 0.03 (0.05) SuccessLag1 0.33 (0.15) Nonviolent:Fem Rights 0.11 (0.04) State Fixed Effects X X X X X Observations 464 399 411 411 408 R 2 0.80 0.83 0.85 0.85 0.58 Adjusted R 2 0.77 0.80 0.82 0.82 0.50 Residual Std. Error 0.20 (df = 396) 0.17 (df = 332) 0.17 (df = 347) 0.17 (df = 346) 0.16 (df = 344) Note: p<0.1; p<0.05; p<0.01

The results of Model 4 demonstrate how the interaction between women s rights and gender diversity influences nonviolent tactic adoption. As can be seen, the model results differ somewhat from the previous models results. While women s rights continue increase the probability of nonviolence, given that the campaign is not gender diverse, this result is no longer statistically significant. Gender diversity continues to have a statistically significant positive effect on nonviolent tactics. However, this effect is only significant when the gender diverse campaign is in a state with lower levels of gender equality. In contrast, while the interaction term is positive, it is not statistically significant. Therefore, it appears that campaign-level gender diversity has the strongest effect on tactical choice in states where levels of women s rights are low. This is a relatively surprising result that counters the theory posited above. However, it is possible that gender diverse campaigns arising in states with low levels of gender equality may be directly challenging the gender norms of that society. Therefore, it is possible that these campaigns are even more likely to embody gender equality norms within the campaign. In which case, the influence of gender diversity may have a more powerful effect in those states with lower levels of gender equality where women s inclusion in the campaign may symbolize a break with previous societal norms than in gender equitable states were these campaigns are conforming with the status quo. Finally, Model 5 demonstrates how nonviolent tactics and state-level gender equality interact to influence the probability that the campaign will successfully achieve its goals. Surprisingly, despite our prediction that nonviolent campaigns in gender equitable states would be more successful than either violent campaigns in gender equitable states or nonviolent campaigns in gender unequal states, the results present a contradictory pattern. While the constituent terms for women s rights and nonviolent tactics are both positive and statistically significant, the interaction term is negative and statistically significant. This implies that while nonviolent methods increase the probability of success in states with low levels of women s rights, they are less successful in states with a strong record of women s rights. Further, the results indicate that violent campaigns in gender equitable states are more likely to be successful than nonviolent campaigns. However, despite the negative influence of gen- 23