Migrate, Cooperate, or Resist: The Civilians’ Dilemma in the Colombian Civil War, 1988–2010

By using innovative microlevel data from Colombia’s civil war, this article examines protests, one of many actions that civilians employ to resists peacefully armed actors during civil war. By using a cost-benefit analysis, the author argues that inhabitants can mobilize and solve their collective action problem when the costs of migrating or supporting one side in a war are too expensive. By examining the civil conflict on Colombia between 1988 and 2010, the author finds that civilians are likely to protest armed actors under three conditions. First, autonomous communities, which enjoy benefits associated with their territory, are likely to protest because migration would be costly; moreover, the institutional strength of these communities can decrease the costs of mobilization against armed actors. Second, foreign aid distributed through nongovernmental associations can empower individuals to take action by providing selective incentives for mobilization, thus decreasing costs. Last, high levels of violence, up to a certain threshold, increase the costs of nonparticipation and induce civilian action.

By using innovative microlevel data from Colombia's civil war, this article examines protests, one of many actions that civilians employ to resists peacefully armed actors during civil war. By using a cost-benefit analysis, the author argues that inhabitants can mobilize and solve their collective action problem when the costs of migrating or supporting one side in a war are too expensive. By examining the civil conflict on Colombia between 1988 and 2010, the author finds that civilians are likely to protest armed actors under three conditions. First, autonomous communities, which enjoy benefits associated with their territory, are likely to protest because migration would be costly; moreover, the institutional strength of these communities can decrease the costs of mobilization against armed actors. Second, foreign aid distributed through nongovernmental associations can empower individuals to take action by providing selective incentives for mobilization, thus decreasing costs. Last, high levels of violence, up to a certain threshold, increase the costs of nonparticipation and induce civilian action.
On January 27, 2005, Autodefensas Unidas de Colombia (AUC) arrived to the hamlet of La Dorada in the municipality of San Miguel, Putumayo, and killed six Ecuadorians. The community demanded peace from paramilitaries in a mobilization held the next day of the massacre. The response from AUC to the protest was brutal. In the following months, paramilitaries threated and killed some of the protest organizers (CINEP 2005). Paramilitaries were no going to tolerate any civilian defiance to their authority.
Mobilizing peacefully against either side of a confrontation, as in the previous example, aims to shape the behavior of armed actors and convince them to stop hurting the community. Regardless of the target of the protest (insurgents, state forces, or both), civilians might be reluctant to participate in mobilization if they enjoy the benefits of the struggle without assuming any cost. Staging public manifestations might trigger a violent response from armed actors, who interpret such actions as expressions of disloyalty (Moreno León 2017). Some civilians might avoid participating and expect others to assume the risks since an unarmed mobilization cannot necessary deter future violent actions but increase their level of vulnerability in the civil conflict. If everyone shares this view, no protests will be held. Thus, the following question emerges: Why do people join a peaceful mobilization at all?
During a civil war, civilians have limited options for survival, and all have costs. They can leave and seek a safer place, but this means selling their possessions and settling in a new and unknown location. They can cooperate with one side, but this might mark them as targets of the other. They can resist armed actors, but this puts resisters and nonresisters at risk, as militias might punish them (Arjona 2016;Barter 2012). After comparing the costs of these options, civilians choose to protest against war only if the price of opposing it is lower than the price of migrating or cooperating.
In this article, I argue that mobilizing against armed actors is the preferred strategy in at least three scenarios in Colombia's civil war. First, a community's autonomy ties individuals to entitled benefits on the territory, making migration costly. Autonomy also provides an available institutional structure that can decrease the costs of mobilization against armed actors. Second, foreign aid distributed through nongovernmental organizations (NGOs) can provide selective incentives for mobilization while empowering individuals to take action, thus decreasing costs. Last, a high level of violence in the territory until a certain threshold increases the costs of nonparticipation, inducing civilian action.

Role of Civilians in Civil Wars
Civilians adapt and adjust in response to the military struggle that may have become part of everyday life (Lubkemann 2008;Barter 2012). The regularity of war forces individuals to make strategic decisions about how to confront the challenges imposed on them. Rebels and civilians are two separate actors with bargaining powers that may not necessarily favor the rebels (Arjona 2016;Mampilly 2011). The strength of community institutions defines civilian bargaining power and thus the level of agreement in negotiations, or in this case, level of cooperation between the two actors. When civilians have strong community organizations, bargaining power favors these civilians, who can make demands on their partners (i.e., the insurgents) such as requiring them to be accountable for their actions. In contrast, when community organizations are weak, bargaining power favors the insurgents, who can make demands on the community (Arjona 2016). In a different situation, organized communities successfully negotiated an agreement with armed actors. Normally, civilians agree to share their resources, as long as armed actors respect their autonomy in handling community relations (e.g., punishing members of the community who refuse to maintain neutrality in the war) (Wilson 1992;Anderson and Wallace 2013;Kaplan 2017).
From this viewpoint, civilians may not be helpless or weak. They are independent actors with their own demands, as long as they are equipped with an advanced level of organization. However, armed actors sometimes underestimate civilians' capacity to organize and attempt to impose their regime without considering the people's preferences, which pushes inhabitants to resist armed actors (Arjona 2016). 1 In this case, civilians might arm themselves and fight back to protect themselves from the abuse of armed actors. For example, Peruvian peasants created self-defense organizations called rondas campesinas to protect their communities from the Shining Path (Fumerton 2001).
In other situations, civilians prefer to use peaceful means to say no to armed actors. Case studies show that civilians conform to autonomous organizations (e.g., Zones of Peace, Communities of Peace) to bargain with armed actors and convince them to moderate their behavior toward civilians (Wilson 1992;EPICA and CHRLA 1993;Avruch and Jose 2007;Valenzuela 2009;Barter 2012;Kaplan 2017;Anderson and Wallace 2013). However, other civilians adopt a different approach. They prefer to hold public demonstrations against one or both parties of the war to show their dissatisfaction (Stoll 1993;Shah 2008). The subject of this article is the nature of these protests.
Regarding the various obstacles civilians face in staging a successful protest, they must first solve the problem of collective action. Citizens might be reluctant to mobilize if they enjoy the benefits of the struggle without assuming any cost. Staging public manifestations might trigger a violent response from armed actors who interpret such actions as expressions of disloyalty (Moreno León 2017). 2 Some civilians might avoid participating, expecting others to assume the risks. If everyone shares this view, no protests will be held. The question then becomes, Why do people mobilize peacefully? In other words, how do they overcome the problem of collective action? 1 Most of the literature analyzes the conditions under which citizens develop specific methods to oppose the government.
Scholars review the characteristics of nonviolent organizations (e.g., institutional) and analyze how these features explain why these movements achieved a change in the status quo (Ackerman and DuVall 2000;Stephan and Chenoweth 2008). Normally, the current literature disregards the possibility that people can come together and mobilize against either the rebels or both conflicting parties. 2 See Moreno León (2017) for further analysis on the effects of protest behavior against armed actors on violent dynamics during civil wars.

The Civilians' Dilemma
When civilians face dangerous situations during civil wars, they must make difficult decisions to survive. The first alternative civilians contemplate is to leave town. Civilians tend to emigrate when security and economic conditions are unbearable to the extent that they foresee a better future in other lands (Davenport Moore and Poe 2003;Engel and Ibañez 2007;Adhikari 2013). 3 The second alternative is to stay and cooperate (in the broadest sense of the word) with one of the armed sides in the confrontation. In this situation, one side might provoke the other to target civilians indiscriminately in an attempt to force them to switch sides (Kalyvas 2006;Humphreys and Weinstein 2008). A third option is to resist the presence of armed actors, which may incite violence from both sides (Wilson 1992;EPICA and CHRLA 1993;Stoll 1993;Shah 2008;Barter 2012;Anderson and Wallace 2013;Kaplan 2017). According to certain conditions, individuals make their choices by ranking their preferences among these options. I argue that three factors influence this decision-making process, all of which involve the problem of collective action. The first factor is the connection between community and territory. In this situation, societies are committed to remaining in their region and protecting their unity despite obstacles so that they can preserve the benefits they are entitled to in their land of residence. The second factor is the relationship between the community and external actors or organizations. External actors can provide selective incentives to local residents to build cohesive communities organized around common goals such as development, peace, and appropriation of the territory. In this situation, foreign actors become agents who subsidize the costs of the collective action problem for the good of the community. Finally, the increased probability of being a target in the war reduces the probability of nonparticipation. When the cost of a passive attitude equals the cost of resisting, individuals may be inclined to mobilize against armed actors. However, when the risks of protesting become extremely high or exceed a risk threshold, people may choose to flee.

Autonomy as a commitment to the territory
People protest armed actors when the expected utility for remaining in the zone is higher than that of migrating. This depends on the cost function, the cost of migrating (C mi ), and the cost of staying and mobilizing against armed actors (C sm ). The latter equals the aggregation of the costs for mobilizing against armed actors (C m ), as well as the costs of repression by other armed actors (C rm ), who are likely to punish them for not supporting their armed group. Therefore, mobilization occurs only if the following equations are satisfied: C mi is greater than C sm when the territory has high cultural and economic value, such as the territories of ethnic groups whose identities are deeply rooted in their homeland. Regarding various factors, landscape shapes identity construction by delineating the economic and social relationships among people (Gordillo 2004;Escobar 2008). In the individual, this dynamic promotes the perception of belonging to a certain region, triggers ethnic solidarity, and embodies the sense of community (Kosek 2006). However, only when these social groups feel threatened by external agents (e.g., dam and mining projects, civil war violence) do they see the necessity of defining the border of their region as a defense mechanism and an opportunity to defend their autonomy (Parajuli 1996;Escobar 2001;Kosek 2006;Öslender 2008). Forced migration then presents itself as a mechanism that dispossesses the territory from its true occupants, putting in danger the transmission of cultural heritage to future generations (Gilbert and Doyle 2011). I discuss how the relationship between an ethnic group and its territory is strategic, as official ethnic and cultural recognition depends on the permanence of this bond in certain collective spaces. Given the fundamental relationship between ethnic identity and territory, I argue that it is expensive for ethnic groups to relocate. 4 The net utility of staying in a territory must be compared both to migration and to staying and cooperating with one side of the conflict. Cooperating with either side of a confrontation must produce a payoff that is lower than that of maintaining the status quo at the expense of popular opposition. In this context, inhabitants mobilize against armed actors only if the costs of doing so (C sm ) are much lower than those of supporting an armed actor (C sc ). C sc is first a function of the risk of violence from the remaining armed actors (C rc ) and second of the resources that inhabitants provide to the armed actor they support (C g ). Therefore: In this context, (C m ) becomes an essential part of the problem. Communities that already have a high level of organization tend to have low transaction costs when deciding to protest. These communities have several strong institutions including community chiefs, religious leaders, and indigenous guards (or selfdefense groups) with sizable memberships (Kaplan 2017). 5 These institutions have a history of facilitating cooperation among members, which is an asset when coordinating protests against armed actors in the community (Ostrom 1999). In addition, they include political entrepreneurs with the ability to coordinate member contributions and punish those who do not participate in the provision of the public good (Taylor 1987). Members of these communities prioritize group and community goals over individual interests and can thus maintain autonomy from armed actors (Lichbach 1998;Petersen 2001).
Given that these communities are highly organized and their identity is closely associated with their territory, their resistance is perceived as credible. The persistence of people staying in the territory, despite the risks, sends an informative signal to armed actors: the community is not going to move away or support any side, despite threats of violence against them. By publicly signaling their compromise, the decision to resist becomes irreversible. This signal has an immense impact on the behavior of armed actors, who avoid punishing members of autonomous communities strongly committed to resistance (Arjona 2016). 6 Repression only exacerbates the hostility toward the armed actor and encourages that actor to support the other side (Kaplan 2017). Therefore, (C rm ) is lower for these communities than for other social groups.
In Colombia, the Indigenous and Afro-Colombian communities have a strong institutional history and power. Colombian law recognizes the collective property rights of these groups and grants them the autonomy to govern their territories under the figures of resguardo or consejo comunitario, respectively. 7 Rappaport (1996) and Wade (2005) explain that this recognition shapes how nonethnic Colombians perceive these groups and enhances the identification of ethnic Colombians, further dividing insiders and outsiders. Therefore, ethnic identity acquires immense value, not only because of deeply rooted emotional ties but also because ethnicity is a strategic tool used by ethnic communities to acquire autonomous political status. Thus, autonomous status facilitates the growth of strong institutions that define the principles and roles of all community members (Hernández Delgado 2004;Ranincheski and Moreno 2010;Kaplan 2017).
Under these circumstances, autonomy, ethnic membership, and property rights are strictly tied to the territory. Consequently, individuals in these communities choose not to migrate, despite the violence of civil war, as doing so means losing benefits entitled to members of the ethnic community in that territory. Each ethnic autonomy then is prone to mobilize both, because the transaction costs of mobilization are low (and in some cases, public demonstration reinforces their identity as a community) and the exit option is too costly or nonexistent. Hence, municipalities that count with several ethnic autonomies in their territory are more likely to stage protests against armed actors. Therefore: H 1 : Municipalities with many autonomous ethnic organizations are more likely to protest against any armed actor of a civil war.

Patrons of peace
Not all communities have the means to proclaim themselves autonomous. The question as to how other nonethnic communities overcome their collective action problem thus emerges. Case studies mention the participation of external actors of the community such as religious communities and NGOs in the creation of nonviolent resistance movements (EPICA and CHRLA 1993;Orjuela 2003;Hilhorst and Van Leeuwen 2005;Anderson and Wallace 2013). However, these analyses neglect to review how outside agents catalyze peaceful demonstrations against armed actors in civil wars.
I argue that external agents such as NGOs are well-endowed political entrepreneurs that can consolidate autonomous and committed organizations. They distribute selective incentives (SI pe ) among members of the community to shape and transform community preferences (Lichbach 1998). In this case, political entrepreneurs use their resources to recruit and retain members of their organizations, rewarding those who contribute to the provision of public goods with private goods (Taylor 1987;Lichbach 1998). For instance, the Peace and Development Corporation from Medium Magdalena is a well-known Colombian NGO that channels foreign aid resources for developing peace-building processes within communities. This organization supports local initiatives to address the structural causes of conflict and empower communities by teaching them about peaceful solutions such as nonviolent resistance (Moreno León 2009; Barreto Henriques 2010).
After the Cold War, the foreign aid system transformed to improve the impact of aid on reducing poverty. The new strategy entailed supporting small-scale projects that the community helped design. Unfortunately, donors were incapable of developing such endeavors, since they lacked the expertise and personnel to conduct these projects. Therefore, they delegated this responsibility to NGOs (Ahmed and Potter 2006;Reimann 2006). NGOs have played an important role in civil conflicts by implementing peace-building policies to boost economic development and democratic governance. They have attempted to rebuild communities by solving part of the social distress that civilians suffered regardless of the outcome of peace negotiations among conflicting parties (Mitchell 2007). Therefore, they developed a series of economic and political projects, trying to mitigate the deteriorating economic conditions of the community and bolstering local organizations to create conditions for lasting peace at the local level (Orjuela 2003;Hilhorst and Van Leeuwen 2005;Alther 2006;Carey 2012;Dosch 2012;Fearon, Humphreys, and Weinstein 2015). Thus, NGOs provide a credible promise on the delivery of private goods to those willing to stay and work for the community. In this scenario, C mi is bigger than C sm , because NGOs provide SI pe that may not be available in another location (see equation 5): . mi m rm pe Foreign aid projects administrated by NGOs convince people to unite and work for the community. The dialogue and cooperation mechanisms embedded within these organizations help solve their collective action problems. Understanding the importance of individual efforts in the consecution of public goods (Lam and Ostrom 2010;Fearon, Humphreys, and Weinstein 2015), communities may continue working together despite the absence of the economic resources promised or delivered by NGOs. The community organizations that NGOs help to build and fortify often shape the conscience of their members to the extent that they become politically active in their communities (Boulding and Gibson 2009;Boulding 2010;Murdie and Bhashin 2011). For example, with the funds of the European Union and support from the Peace and Development Corporation from Medium Magdalena, the small population of Micoahumado, Colombia, created Humanitarian Spaces. They then declared themselves a third party in the war, engaged in a series of negotiations with local armed actors, and demanded better behavior towards civilians during their presence in the territory. This project immediately produced backlash from the government such that national authorities temporarily froze all foreign aid funds to the region. Despite this obstacle, the community continued its struggle (Moreno León 2009). Thus, the intervention of NGOs distributes private goods, SI pe , and reduces the costs of mobilization, C m , as they help organize the community: Last, NGOs also provide an invisible cloak of protection to the communities they work with. Armed actors, who are concerned about their international reputation, may find it too costly to punish civilians who demand that their rights be respected (Eguren and Mahony 1997;Mampilly 2011). 8 International audience costs are likely to deter violence by armed actors, as the global network in which NGOs are embedded is likely to disapprove of violent actions against unarmed civilians, pushing governments to punish those who hurt civilians. Consequently, communities funded by foreign aid resources and supported by NGOs assume lower C rm , given that armed actors are likely to find repression and indiscriminate violence against civilians a costly strategy: In summary, NGOs can increase community cohesion and security, which encourages peaceful mobilizations against armed actors (Arjona 2016). Indeed, these organizations have been decisive in the constitution of many autonomous territories that claimed neutrality in civil conflicts, such as the zones of peace in the Philippines and the Peace Community at San José de Apartadó in Colombia (Orjuela 2003;Alther 2006;Avruch and Jose 2007;Valenzuela 2009). Therefore: H 2 : Municipalities receiving high amounts of foreign aid are more likely to protest against any armed actor of a civil war.

Violence, tolerance, and migration
Thus far, I have noted how certain communities overcame their collective action problems with the organization and distribution of selective incentives. I now evaluate how the use of violence by military organizations affects civilian behavior. The social movement literature tends to assume that government uses repression to discourage collective action by increasing the costs of mobilization (Opp and Roehl 1990). However, excessive use of coercion against a population might be costly for the authorities as well.
For example, this behavior might divide the ruling elite (Stephan and Chenoweth 2008), decrease the legitimacy of the government (Opp and Roehl 1990), or push the population toward indifference where staying at home and mobilizing against the government no longer matter (Tarrow 1998). However, deterrence by repression may be less applicable in civil wars. While civilians may have the opportunity to abstain from demonstrating when enjoying the benefits of social mobilization in unarmed conflicts without assuming any costs (e.g., repression), the situation is different when civilians live in contexts where sovereignty is fragmented and the boundaries between the sides in the confrontation blurred. In such cases, nonparticipation can be more lethal than being an active agent during armed civil conflicts (Kalyvas and Kocher 2007). Winning the war partly depends on sustaining the population's active support. Civilians provide information, economic resources, and shelter. Thus, armed actors find it necessary to punish and deter civilians from cooperating with the enemy. However, combatants know that the efficiency of such methods hinges on the effective discrimination of the population between friends and foes, as mistakes in the selection process sever the loyalty of the population (Kalyvas 2006). Therefore, armed actors tend to trust local informants to reveal the supporters of the enemy. While the levels of violence are low and the screening mechanisms works efficiently, civilians will have no incentive to mobilize against armed actors. However, screening processes may fail, as informants often provide false information that benefits their private agendas (e.g., marital conflicts or revenge) (Kalyvas 2006). Despite armed actors' use of selective screening to punish traitors, passive civilians may be victimized (Kalyvas and Kocher 2007). This increases the costs of nonparticipation for civilians who may be targeted regardless of guilt when violence escalates. Thus, participation in this context becomes a rational choice.
In this case, how does repression shape the preferences of civilians? I argue that the willingness of the population to protest against armed actors increases when the possibility of being victimized is high enough such that being passive becomes more costly than mobilization. Case studies evidence how the escalation of civilian abuse by armed actors facilitates the surge of civilian organizations initiated to oppose insurgents and state forces. These organizations offer better economic and security conditions than the alternatives (e.g., emigration, support for one of the combatants in a conflict) (Wilson 1992;EPICA and CHRLA 1993;Avruch and Jose 2007;Valenzuela 2009;Barter 2012;Anderson and Wallace 2013;Kaplan 2017). Indeed, Arjona (2016) argues that communities tend to resist when a military organization imposes its own regime that disregards the preferences of civilians. In this case, civilians might interpret the continuous use of violence as an attempt to destroy their social order. Therefore, an increase in repression exerted by armed actors can incentivize the mobilization of civilians against armed parties of the confrontation. Civilians may then perceive indifference as a suboptimal strategy, instead taking the opportunity to publicly show their preferences to persuade armed actors to change their behavior towards the community.
However, this relationship is not linear, as a high level of violence can deter the mobilization of the population. When repression is widespread in the territory, individuals may believe that changing the course of the war is futile. Assuming additional costs in this case, such as protesting against armed actors, is expected to trigger retaliation against those who organize and mobilize against combatants. In this context, civilians must choose between staying with the hope that the combatants will not bother them, or migrating to a safer place, where they can start over.
In both situations, civilians face risks. Staying is a good option only if the probability of becoming a target is low. In that case, civilians might buy protection from a belligerent group. Unfortunately, their guardian in fragmented zones is incapable of fully guaranteeing their survival (Kalyvas 2006). In contrast, civilians might consider migrating, which they opt to do when they know the new place ensures better security and economic conditions (Davenport Moore and Poe 2003;Adhikari 2013). When the level of violence is low, this uncertainty might deter people from migrating (Engel and Ibañez 2007). However, the lack of information is dissipated when the level of violence increases. When repression increases, the first civilians to abandon the town are likely to spread news on the conditions of their new secure environment to those who remain (Adhikari 2013). Therefore, civilians who have not fled tend to update their beliefs on the costs of staying, which increases the number of those who migrate, slowly creating migration herding behavior (Engel and Ibañez 2007). 9 Therefore: H 3 : Protesting against armed actors is more likely when repression costs increase, but when the level of violence is high, the likelihood of protesting against armed actors decreases.

Colombia as a Case Study
In the mid-1960s, the Fuerzas Armadas Revolucionarias de Colombia (FARC) and the Ejército Nacional de Liberación (ELN), two of the longest-lived guerrilla groups of the Western Hemisphere, started their struggle against the government. At the beginning of the confrontation, both guerrillas were small groups that did not represent a real threat to the government since they operated in remote areas (González et al. 2002). In the early 1980s, such tendency changed. Both guerrillas multiplied their military fronts and their economic sources (e.g., extortion, kidnapping, coca crops) with the only purpose of taking control over the government. While insurgents were gaining more strength, local elites (some linked to drug trafficking) constituted a series of paramilitary organizations to fight insurgents and secure their control over different illegal economies (González et al. 2002). Besides the countless lives lost during the past decades, the civil conflict generated nearly seven million internal displaced people between 1988 and 2010. 10 The long duration of the Colombian civil conflict has driven civilians to consider the war as part of daily life (Pécaut 2001). As such, civilians have learned to survive while caught in the cross fire (Lubkemann 2008). In addition to autonomous organizations that encourage dialogue between civilians and armed actors (e.g., Comunidad de Paz de San José de Apartadó) (Hernandez Delgado 2004, Valenzuela 2009, Kaplan 2017, civilians also implement public demonstrations against insurgents, paramilitaries, and state forces to shape combatants' preferences for violence. According to the Social Struggles Dataset from the Centre for Research and Popular Education (CINEP), 19.9 percent of the total number of demonstrations in Colombia between 1988 and 2010 were protests against one or more armed actors in the war. In the late 1980s, mobilizations against armed actors were not so frequent. However, such trend shifted in the mid-1990s. Protest behavior rise as well as the recrudescence of the civil conflict. According to Granada and colleagues (2009), the number of civilian causalities, kidnappings, and internal displaced people increased dramatically between the end and the beginning of the twenty-first century. In this sense, protests against armed actors is a reflection of the hardest moments of the Colombian civil conflict.
This article attempts to explain how civilians make decisions regarding the local dynamics of the war. I compare how municipalities' political and military characteristics influence protest behavior against combatants. Unfortunately, little cross-national information is available on the local dynamics of civil war and protest behavior (Rustad et al. 2011). 11 Furthermore, social mobilization data are not typically aggregated by town and focus solely on antigovernment protests. However, in Colombia, private and public organizations have made efforts to collect local-level information about the dynamics of the civil war and protest behavior (Restrepo, Spagat, and Vargas 2006). As a result, Colombia presents a good case for testing the conditions under which civilians in civil conflicts resist violent organizations.

Research Design
The dependent variable is number of protests against any armed actor in each town per year. I employed information extracted from the CINEP Social Struggles Dataset. This data set uses national and regional newspapers, interviews with social movement leaders, and reports from social organizations to collect information about protests, which it defines as any public demonstration with at least ten participants. From that data set, I selected all civilian protests against guerrillas, paramilitaries, state forces, or armed actors between 1988 and 2010. 12 I tested the first hypothesis by creating the variable autonomy, which measures the number of Indigenous and Afro-Colombian autonomous communities the government recognizes in each town. I used the registers from the Interior Ministry of Colombia.
I evaluated the second hypothesis according to the amount of foreign aid each town receives per year and created a variable named foreign aid to account for the number of official development assistance (ODA) projects each town implements per year. This variable was developed based on the reports of the Presidential Agency for International Cooperation of Colombia.
Finally, to test the last hypothesis, I employed the rate of victims killed and wounded by armed actors, and created the variable victimization rate using the number of civilians harmed by armed actors. 13 This was derived from the Human Rights and International Law Observatory data set and population projection data from DANE. I then squared this variable to account for the negative effect of extreme violence on the probability of mobilization against armed actors.
Furthermore, I employed four control variables. First, I extracted the number of Catholic and Christian churches in the municipality in 1995 using data from the Social Foundation. Using this variable, I evaluated the role of religious communities as external agents in the promotion of organizational processes in the community in favor of peace-building initiatives (Wilson 1992;Kaplan 2017). Second, I used a dummy variable to evaluate the impact of regional capitals on mobilization. Political centers are more likely to have resources and the community relationships required for mobilization. This variable equals 1 when the city is the regional capital and 0 otherwise. Third, I employed population density to assess how easy it was for the population to organize a demonstration. I formulated this variable by using population projection data from DANE and geographic area data from the Geographic Institute Agustin Codazzi. Because of the variance of this variable, it was logged. I expect that mobilization is easier in populous communities. Fourth, I included in the model the Unsatisfied Basic Needs Index of the Colombian Census of 2005 to address the effect of the level of economic development on the probability to protest. Finally, I accounted for the number of protests unrelated to war, using this measure to assess the ease with which towns reached the minimum threshold for mobilization (Petersen 2001).
I estimated a set of zero-inflated negative binomial models to evaluate the effect of the independent variables on the number of protests against armed actors. I use this model because the number of protests against any armed actor is a discrete variable with nonnegative integer responses. Furthermore, as it is rare to protest against armed actors, the dependent variable has many zeros. Potentially, the creation of zeros responds to two processes: there might be towns in which inhabitants never stage protests and cities in which civilians have the opportunity for public demonstrations but choose not to protest. The selected model not only address the characteristics of my discrete response of counts but also helps adjust the estimates of the count equation according to the probability of finding zeros in the sample (Long 1997). In the inflated equation, I used only the control variables. In addition, I clustered the standard errors by municipality in all the models. The unit of analysis was municipality-year.

Results
The empirical analysis supports the first hypothesis. Municipalities with many ethnic autonomous organizations are more likely to protest all sides of the civil war. The estimations demonstrate a positive and stable correlation between Indigenous and Afro-Colombian autonomies and the number of protests against the civil war. In fact, an increase in the number of autonomies enhances the expected number of protests by 1.9 percent when all other variables are held constant (see Figure 1). 14 Therefore, the provision of particular rights attached to the territory provides incentives to ethnic communities to remain in the territory despite military confrontation and to mobilize against the war. 15 The second hypothesis argues that municipalities receiving a high amount of foreign aid are more likely to protest the civil war. The findings confirm that an increase in the number of foreign-aid projects in a municipality increases the chances of mobilization against combatants. According to the results of the estimation, a unit increase in foreign-aid projects boosts the expected number of protests by 4.4 percent when all other variables are held constant. In this case, the participation of foreign actors can convince the population to remain in the territory and mobilize to induce better treatment from armed actors. Thus far, I have accounted for the propelling effect of NGOs through ODA on the number of protests against armed actors. Other external agents such as religious organizations might produce a similar effect on civilians' mobilization. My empirical analysis revealed that a unit increase in the number of Catholic and Christian churches decreases the probability of having no protests against armed actors by 21.5 percent. Despite the fact that this variable is insignificant in the count equation, the result in the inflated equation suggests that religious organizations might influence the probability of having a protest but not the quantity thereof.
The last hypothesis argues that an increase in the number of victims increases the number of protests against armed actors. However, I also contend that there are fewer protests when the level of violence is too extreme. According to the estimation, people tend to mobilize more often when facing violence and when the probability of being targeted is high. Figure 2 shows how a rise in violence against civilians by armed 14 The statistical analysis also show that municipalities embedded with experienced and few ethnic autonomies are more prone to mobilize against armed actors than municipalities equipped with several ethnic autonomies without mobilization experience. See the online appendix for further analysis, as well as discussion on the impact of cooperation among ethnic autonomies on protest behavior. 15 Until now, I have been interested in reviewing the effect of well-organized communities with strong attachments to the territory.
Based on the empirical analysis, municipalities that mobilize more frequently around issues not related to the civil conflict are more likely to protest against armed actors. In fact, a unit increase in the number of protests unrelated to the war boosts the expected number of protests by 3.9 percent when all other variables are held constant. actors increases the incidence rate ratio of a protest in the territory. For example, municipalities in which 5 percent of the population was victimized protested 14 percent more often than did municipalities where only 4 percent of the population was victimized. However, when the rate totals 7 percent of the population, the incidence rate ratio decreases when all other variables are held constant. 16 For instance, towns where 12 percent of the population was victimized protested 5 percent less than did towns where 11 percent of the population was victimized. Therefore, violence against civilians can motivate mobilization against military organizations, but only if the benefits justify the costs. For example, people do not mobilize when the level of violence reaches a certain threshold (i.e., cost) that prevents any change in the outcome (i.e., benefit). 17

Robustness Checks
One concern is that the dependent variable fuses all demonstrations, disregarding their target. Therefore, I also tested the three hypotheses using a multivariate probit model. I input three equations in the estimation. The first two equations evaluated the conditions under which civilians protest against one side-insurgents, or state forces or paramilitaries-in the war. The third equation assesses the circumstances that drive civilians to mobilize as unbiased actors and demand from both armed actors better behavior toward the community. As shown in the online appendix, ethnic autonomies and foreign aid still explain why people protest armed actors. Of course, the probability of having a protest change according to the target exists, but both variables demonstrated a strong and positive relationship with protest behavior against armed actors. 18 16 The marginal effect of the victimization rate is insignificant when around 6.8 percent and 7.6 percent of people are hurt in civil conflict. Specifically, after such fragmentation, the probability of an additional protest against armed actors declines. 17 I estimate the full model with two additional variables that address the size of the population and area of the municipalities. As described in the online appendix, the main independent variables still demonstrate a strong correlation with civilian mobilization. I also estimate the full model only with the municipalities that have experienced violence from armed actors. As described in the appendix, the main independent variables still demonstrate a strong correlation with civilian mobilization. Finally, I estimate other nonlinear models with the same results found in the appendix. 18 Despite the positive correlation between ethnic autonomies and protest against both armed actors, the estimate is not different from 0. Nevertheless, as the model suggests, ethnic autonomies can mobilize against insurgents or state forces. Again, the rate of victims hurt by armed actors produces an inverted U shape for the probability of a protest against any armed actor. Consequently, the three hypotheses postulated in this article explain the minimum conditions that drive civilians to overcome their collective action problem and mobilize against armed actors despite the target of the protest. 19 Thus far, I have assumed that civilians are more likely to mobilize against armed actors when the costs of migrating are too high. However, is this assumption valid? I tested the impact of the main independent variables that predict protest behavior against armed actors given the possible choice of migration. In this case, I estimated a set of multivariate probit models that demonstrate the correlation among the errors of two or more seemingly unrelated equations (Davidson and MacKinnon 2004). I input three equations into the models. The first equation evaluates the conditions under which civilians protest armed actors, the second tests the circumstances under which citizens forcibly emigrate, and the third assesses the factors that explain the arrival of internal displaced people to municipalities in Colombia.
As shown in the online appendix, the three hypotheses explain protest behavior against armed actors. However, interesting results regarding forced migration processes in civil war emerged, which require further analysis in the future. Ethnic autonomies are positively associated with immigration and emigration processes in Colombia. Some studies emphasize that ethnic autonomies in Colombia form a network that shares information and supports fellow ethnic groups. When armed actors execute violent actions against the population, ethnic autonomies warn the ethnic communities in the neighborhood to prepare for the arrival of the armed actor. Ethnic autonomies then preemptively migrate to other places as a mechanism to defend against possible attacks on their members. They temporarily abandon their territories, relocating to places where they will be supported by other ethnic organizations, and eventually return to their territories (Öslender 2008;Castillo Valencia 2009). For ethnic autonomies, mobilization and migration are complementary strategies used to increase their probability of survival. In contrast, foreign aid cannot predict migration processes in Colombia. Despite the fact that NGOs distribute foreign-aid resources to the community, stronger factors shape the preferences of civilians for migration, such as violence. Finally, escalation in the level of violence increases the probability of the emergence of internally displaced people. However, when the rate of victims is 2 percent of the population, the probability of having an additional emigrating person decreases, probably because the most cost-sensitive citizens have already left. In contrast, when the level of violence escalates, the probability of protesting against armed actors also increases until the percentage of victims totals 3 percent of the population, after which the probability of a protest against armed actors declines. These results suggest that the first set of people who abandon the town are less committed to the community, while those who remain find that mobilizing can be an effective mechanism by which to improve the security of the town. This preliminary evaluation indicates that it is necessary to contemplate the conditions under which emigration and mobilizations against armed actors are complementary choices for civilians rather than substitutes. As such, future research should explore the interaction between protest behavior and emigration.
Some scholars argue that foreign aid is not randomly assigned. International cooperation agencies might prefer to assign their resources to strong grassroots organizations. In this case, the level of organization of the communities for a protest against armed actors might be mediated by foreign aid. In this sense, the impact of foreign aid on protest behavior might be a product of the level of organization located in the municipalities. Following this strategy, I could more precisely evaluate the underlying causal mechanisms between level of organization of the communities, foreign aid, and protests against armed actors.
Using protests unrelated to the war as a proxy for level of organization of the municipality, the results show that the variable protests unrelated to the war directly and indirectly affect the probability of a protest in the municipality. The number of protests unrelated to the war has an impact on protest behavior against armed actors through foreign aid. However, this impact accounts for only 1 percent of the total effect of protests unrelated to the war on mobilization against armed actors. Therefore, most of the effect of foreign aid on protest behavior stems from the direct effects explained in this article. 20 19 Although some literature affirms that state forces and paramilitaries often collude in their struggle against insurgents, others contend that paramilitaries are an independent actor. Therefore, I ran another multivariate probit model. The first three equations evaluate the conditions under which noncombatants protest against one side-insurgents, state forces, or paramilitaries-in the war. The third equation assesses the circumstances that drive civilians to mobilize as unbiased actors who demand from all armed actors better behavior toward the community. As shown in the online appendix, the results do not vary greatly from those in the previous trivariate probit model. The main difference appears around foreign aid. This variable effectively predicted mobilization against all combatants except state forces. This situation might suggest that foreign aid intervenes in the selection of the target of the protests, not necessarily discouraging protests but being less enthusiastic about them. 20 For further analysis, please review the online appendix.
Finally, I argue that through the distribution of foreign aid, NGOs can catalyze protests against armed actors. However, previous studies highlighted a positive effect on the level of conflict behavior. In this case, the effect of foreign aid on protests against armed actors might be mediated by the victimization rate. Because foreign aid might escalate violence in municipalities, foreign aid could indirectly impact mobilization, given its effect on the level of violence. Therefore, I estimated a causal mediation model (Imai et al. 2011). Following this strategy, I could more precisely evaluate the causal mechanisms underlying foreign aid and protests against armed actors.
As shown in the online appendix, the results of the estimation indicate that foreign aid directly and indirectly affects the probability of a protest in the municipality. The allocation of foreign aid in the municipality then impacts protest behavior through the victimization rate. However, this impact represents only 2 percent of the total effect of foreign aid on mobilization against armed actors. Therefore, most of the effect of foreign aid on protest behavior stems from the direct effects explained in this article.

Conclusion
Civilians are important agents in the dynamics of civil wars, as they are capable of bargaining with armed actors and vocalizing their preferences. Sometimes, military organizations may determine the order of things individually, and civilians may not necessarily agree. Thus, they may choose to develop a nonviolent type of resistance against the presence of combatants (Arjona 2016). For civilians to conduct such costly actions, they must solve their collective action problem. Some choose to pay the considerable cost of mobilization to obtain small benefits (Wood 2003). Some overcome their mobilization problems and protest against one or all sides of a confrontation. This article has attempted to explain the conditions leading to that decision.
I have argued that people make choices based on the possible costs of their actions. In other words, individuals resist only if the costs of doing so do not exceed the costs of alternatives (e.g., migrating to a safer place or cooperating with one of the armed groups). However, communities and war dynamics have different attributes that influence how the costs are defined during civil wars. These attributes can either reduce the costs of staying and mobilizing against the war, or they can raise the costs of being a passive actor in the territory. In this article, I studied three of these costs: ethnic autonomies, foreign aid distributed through NGOs, and the level of violence.
First, I argued that autonomy among ethnic communities affects their choices. These communities assume elevated costs if they move to other locations. By migrating, they lose part of their identity and the political rights entitled to them in their home territory. Furthermore, the institutions that help autonomous communities function have the means to reduce the costs of mobilization and sometimes, these institutions can empower individuals in the community. As my findings show, these communities tend to mobilize more often than nonautonomous communities against the civil war.
Second, I evaluated the role of external actors by focusing on NGOs as external entrepreneurs. As wellendowed actors, NGOs can distribute selective incentives in communities for two main purposes: to empower local inhabitants and thus promote and strengthen political and communitarian bonds, and to convince individuals to remain in their territory and play a role in its future. The results of the empirical analysis confirm the strong correlation between number of protests and number of foreign-aid projects within municipalities.
Finally, the violence generated from the conflict can alter how the population evaluates the costs of mobilization. When the costs of violence are the same as the costs of mobilizing, inhabitants are willing to protest against the war. However, when the level of violence is excessively high, the population is less inclined to mobilize against armed actors. The evidence has pointed to how the population studied tends to mobilize when the violence produced by the conflict increases. In contrast, when the victimization rate totals around 7 percent of the population, the probability of protests declines. Nevertheless, further research should evaluate how individual civilians survive in contexts of indiscriminate violence following other strategies (e.g., migration) and how the variation in the violent repertoires might influence the probability of having a protest against armed actors.
In this article, I have developed my argument that civilians are rational actors. According to a cost-benefit analysis, civilians choose to protest against armed actors when the costs of such action are less expensive than the alternatives (migrating or supporting one party in the war). Sometimes, people mobilize in favor of a cause despite the lack of benefits of that endeavor. At first glance, self-respect and defiance can drive human behavior toward foolish actions in dangerous contexts (Wood 2003). However, people's conduct cannot be interpreted as irrational. In this case, civilians mobilize because their cost-benefit analysis is formulated under the premise of their moral duty to themselves and the community (Varshney 2003).
Since the behavior of civilians is driven by the pleasure of agency instead of the future benefits of the action, these civilians are less sensible to the costs of mobilization (e.g., repression). In this framework, autonomous communities and organizations supported by NGOs are more likely to mobilize. Both groups count on institutional strength to solve the collective action problem and protest against armed actors. In contrast, the level of violence can pacify demonstrations against the parties at war. Civilians are more sensitive to a loss of human lives in their communities and will mobilize more to demand better behavior from combatants. However, civilians might be considered risk seekers from this perspective, and they will realize that in extreme levels of violence, few people have their same combative spirit and prefer to refrain from contentious behavior.
Much work is needed to understand civilian resistance against combatants during civil wars. The theory in this article explored a civilians' cost-benefit analysis. Future work should evaluate the strategic interaction of both civilians and armed actors. Studying the many contexts in which civilians and armed actors interact and the outcomes of these interactions can improve understanding of civil wars. For example, future research should differentiate the targets of protests. Thus far, my analysis aggregated all types of protests against armed actors, regardless of target. If civilians are strategic, they select a target of mobilization to maximize their benefits and minimize their costs. How do armed actors behave in return? In equilibrium, neither civilians nor armed actors should have the incentive to change behavior according to behavior of the other. Understanding this strategic relationship can strengthen our comprehension of long-term conflicts such as that in the Colombian case and the effects of such conflicts on local communities.

Additional File
The additional file for this article can be found as follows: • Appendix. Descriptive statistics, main models, additional models, and maps. DOI: https://doi.org/ 10.25222/larr.640.s1