Social Responses to Criminal Governance in Rio de Janeiro, Belo Horizonte, Kingston, and Medellín

Over the past generation Latin America has experienced high levels of criminal violence associated with extortion, the drug trade, and other criminal rackets. While there has been considerable research into the role of state policy in controlling criminal violence, there has been considerably less investment in analyzing the success of social action to control both police and criminal violence in the high-conflict zones where criminal groups operate. The inadequacy of the existing literature emerges at least in part from the limited data on successful social efforts to control violence in urban areas controlled by organized crime groups. Drawing on over three hundred qualitative interviews conducted in four cities in Latin America and the Caribbean over an extended period of time, this article examines the various strategies used by civic groups to control violent activities on the part of police and criminal groups. The article observes six different forms of social responses to violence and points to the conditions under which these actions emerge.

Over the past generation Latin America has experienced high levels of criminal violence associated with extortion, the drug trade, and other criminal rackets. While there has been considerable research into the role of state policy in controlling criminal violence, there has been considerably less investment in analyzing the success of social action to control both police and criminal violence in the high-conflict zones where criminal groups operate. The inadequacy of the existing literature emerges at least in part from the limited data on successful social efforts to control violence in urban areas controlled by organized crime groups. Drawing on over three hundred qualitative interviews conducted in four cities in Latin America and the Caribbean over an extended period of time, this article examines the various strategies used by civic groups to control violent activities on the part of police and criminal groups. The article observes six different forms of social responses to violence and points to the conditions under which these actions emerge.
Over the past two generations Latin America has seen a dramatic increase in criminal violence. Much of the recent literature on bloodshed in the region has focused on how governments can help to bring this violence under control (Ungar 2011;Pinheiro 1997, 201-203;Pinheiro 1999;Pereira 2008;Lessing 2017). These strategies, however, often encounter insurmountable challenges in neighborhoods where gangs or other armed actors are present because of the crime-state alliances that emerge from organized illicit activities (Arias 2006(Arias , 2017. Effective crime control in these areas depends not just on state efforts but also on civic action to ensure state accountability, empower reformers, and enlist criminal organizations themselves in reducing violence. Here I examine how civic groups can respond to violence amid the different types of criminality, state violence, and collusion between criminals and the state that are common in Latin America and the Caribbean. Literature focused on North America presupposes the existence of a functioning state that seeks, at times imperfectly, to repress criminal activity, rather than collaborating with criminals in order to advance state actors' political and economic projects, as is the case in much of Latin America and the Caribbean (Arias and Goldstein 2010). Through understanding the nature of criminal violence in a particular locale, this article provides a model for the types of responses to violence that emerge under different circumstances.
The data presented here provide three key insights into the forms of civic response to violence in areas of dense criminal activity. First, civic response is driven by the underlying forms of localized governance resulting from state-criminal interactions that exist in a particular neighborhood. Second, the particularities of national and municipal politics generate opportunities for civic mobilization in the context of neighborhood-specific criminal governance structures. Finally, I will show that civic groups often seek to negotiate with criminals rather than police to address local violence. This can complicate long-term violence control since state action is usually required for durable solutions. Building on data gathered over twenty years in Rio de Janeiro and Belo Horizonte, Brazil; Medellín, Colombia; and Kingston, Jamaica, I offer a layered typology of civic responses to crime and outline the conditions under which these responses emerge.

Criminal Governance and Civic Responses to Crime
Much of the literature on civic responses to crime in Latin America has focused on broader national or urban-level reform efforts (Davis and Denyer Willis 2013; also see Moncada 2016a andBonner 2014). The role of civic groups in containing violent crime in Latin America within crime-affected neighborhoods has received relatively little attention (for exceptions see Moncada 2009;Arias 2004Arias , 2006). Focusing on the neighborhood level, scholars working on criminal ecologies in North America argue that the presence of civic groups capable of establishing local norms and collaborating with police can reduce crime in the neighborhoods where they operate (Bursik 1999, 85-97;Sampson 2012). From a policy perspective, Kennedy (2011Kennedy ( , 2009) has argued that local leaders can help control crime by establishing norms that reject certain behaviors.
In the Latin American context, often characterized by gangs capable of confronting state forces (Durán-Martínez 2015), building norms to repudiate illicit behavior is a more difficult endeavor. Davis and Denyer Willis point out that these approaches face serious challenges to success due to vigilantism, criminal networks, state corruption, and pro mano dura mobilizations that tend to aggravate crime and undermine effective responses (Davis and Denyer Willis 2013). This is particularly the case in cities in Colombia, Central America, Mexico, and Brazil, where armed actors often establish state-supported territorial control. Under these circumstances, armed actors are a driving force in local politics and in shaping civil society (Arias 2017). Independent civic groups might exist in such communities, but given the resources available to armed actors and their ties to the state, civic leaders have little hope of shifting local norms (Arias 2014). Despite these challenges, civic groups do have some space to manipulate norms and engage in other efforts to contain armed groups' most abusive behaviors (Arias 2004;Moncada 2009). The nature of criminal governance in a locale, however, constrains the activities of civic groups (Arias 2017).
The state is undertheorized in North American ecological approaches, which take certain norms of state institutional practices for granted. In Latin America this is not the case. Here there is ample evidence of police and other state actors engaging with criminals in varied ways to shape violent practices so as to advance either individual or collective economic and political projects (Arias and Goldstein 2010;Arias 2017). Relationships between state actors, including police as well as other elected and appointed officials, play an important role in shaping the type of criminal activity that exists in a particular neighborhood. For the purposes of this article, the state is a set of formally empowered actors, including police as well as elected and appointed officials, that are in the employ of public institutions and that have the authority to set and carry out policy. While a variety of state officials can shape relationships with criminals, police often play a critical role since they are charged with primary law enforcement activities. Their various forms of engagement with criminals and the wider community both control and generate violence. Given this critical role, police are an important target and instrument of efforts to control violence.
The impact of nonstate armed actors on contention and protest in Latin America has not received the theoretical attention it deserves. Tilly and Tarrow (2015), for example, argue that states define the limits of contentious politics, but they pay scant attention to the role that violent nonstate actors play in limiting and shaping different forms of protest or contention in the areas they control. Indeed, in their broader discussion of violence by these actors, Tilly and Tarrow (2015, 111, 169-190) focus on how state structures generate or repress different forms of large-scale violent mobilizations but do not discuss how nonstate armed actors, in turn, generate their own substate opportunity structures in the areas they control. Scott's (2009) work on nonstate rule in upland Southeast Asia similarly focuses on how nonstate actors resist state control rather than on how those same nonstate actors shape political mobilization in the areas where they operate. Similarly, Holston's (2008) discussion of the politics of drug gangs in Brazilian shantytowns examines the protest interactions between drug gangs and the state as an example of democratizing dynamics in Brazilian society but does not discuss how those same gangs shape and repress protests against gang activities and other policing strategies in the areas that gangs dominate.
Still, there is some wider evidence that armed nonstate actors often seek to limit protests against themselves or the state (see for example Blok 1974; also see Auyero 2007). Staniland (2012) has pointed out that interactions between armed state and nonstate actors in civil conflicts generate distinct political dynamics. Arjona (2016) has discussed the interactions between civic actors and Colombian guerillas in areas controlled by guerillas. I have shown how armed actors affect civic mobilizations in a Brazilian shantytown (Arias 2014).
In a recent monograph I offer a four-part framework for criminal governance based on the degree of criminal consolidation and the varied nature of relations between armed actors and the state that can emerge in the region. These armed governance contexts shape the pathways available to civic groups to contain armed actors based on civic groups' relative independence and their ability to negotiate with state and criminal actors. The types of violence faced under particular criminal governance conditions and the relationships among civic groups, various state actors including the police, and criminal groups generate opportunity structures that shape how civic leaders can respond to violence (Arias 2006; on opportunity structures more broadly see Meyer and Staggenborg 1996, 1633-1635; on broader competitive conditions and mobilization dynamics see Trejo 2009, 340). This article will develop in more detail how local armed actor structures affect civic mobilization to control violence.

The Normative Context and Opportunity Structure of Civic Action against Violence
The relationships of three types of political actors-civic groups, armed organizations, and state officials-establish pathways for civic responses to violence. Key to violence control efforts are the neighborhood-level grassroots organizations that I refer to generally in this article as civic groups, and short-term movements of residents, often led by actors from those same civic groups. Civic groups and residents of gang-controlled areas have complex relations with armed actors, who provide services, are part of local friendship and familial networks, and contribute to some families' survival strategies (see Gay 2005) but who also contribute to local violence and hardship. For the most part, local leaders would prefer not to undertake actions that could put them at odds with police or armed groups.
Simultaneously the local population constrains criminals. While they have the capacity to exercise violence, criminals also depend on the population for protection from the state and other criminals. Open protest against criminals can attract state attention that interferes with illicit business activities or leads to arrest. These constraints similarly limit criminal violence since confrontations can erode support over time.
The state also plays important roles in defining the nature of the opportunity structure for antiviolence collective action. The state is a complex entity that contains many different and often competing actors. Two sets of actors are key to this article: (1) police, and (2) elected officials and high-level appointees. Police play an essential role in creating and responding to security conditions by determining and carrying out crime control policies. Further, their relationships with neighborhood citizens, civic leaders, and often criminals affect local security and can be constructively engaged by citizens in efforts to control violence. Their contacts with criminals can produce and reinforce violence. As a result, concerns about police in a locality often provoke collective responses that affect violence and relations among the state, social actors, and criminals. Police violence and corruption are inextricably linked with crime, and responses to police activities are as much concerned with constraining police violence and corruption as they are with compelling police to control crime. The power of police is balanced by elected and appointed officials, who also play critical roles in generating security. These actors can also choose to collaborate with criminals in seeking to produce their desired security, economic, or political outcomes and, indeed, the degree of collaboration between the state and criminals is driven by the ways that police and politicians interact with criminals. At the same time, officials can provide structures and networks with which citizens can interact as they seek to contain both police and criminal violence.
The exchanges among these actors and the types of violence faced under armed governance structures shape distinct violence control strategies. Under criminal disorder, violence stems principally from conflict among disorganized gangs. Since disorganized gangs exercise little civic control, social leaders have various options to contain conflict. Under divided governance, civic groups, which are often subject to armed actor control, have much less operational space. In these cases, civic groups often target state policy in the areas where they operate while seeking to work constructively with criminals. Under collaborative governance, active collusion of organized crime and the state substantially limits civic groups' operational space. Here, civic groups seek to modulate criminal and state behavior with criminal support. Finally, under tiered governance state actors help modulate criminal behavior. In these circumstances civic actors collaborate with state and criminal actors to reduce violence. Table 1 illustrates the types of strategies that may emerge under different types of criminal governance structures.

Cases and history
The data supporting this argument draw on research conducted over an extended period in Rio de Janeiro and Belo Horizonte, Brazil; Medellín, Colombia; and Kingston, Jamaica. Rio, Medellín, and Kingston were chosen for research because of a history of criminal violence. All have strongly territorial criminal gangs and share a great deal in common with cities in Colombia, Venezuela, Mexico, and northern Central America. Belo Horizonte has moderate violence and gangs exercise more limited territorial control. I conducted research there to examine policy innovations and to provide a Brazilian counterpoint to the work that I had conducted in Rio in the 1990s. Data were gathered in the four cities through participant observation and interviews in poor and working-class neighborhoods that were often dominated by criminal gangs. Much of my research has focused on the grassroots organizations that operate in these neighborhoods, which include neighborhood associations and other social organizations.
Each of the individual examples discussed provides independent insights into how citizens can effectively respond to crime. At the same time, each exists within a particular national context in which trajectories of violence and wider political dynamics affect civic opportunity structures. These crime and policing conditions generate certain types of civic responses and constrain the options available to civic actors in the political context in which they operate.
In Colombia, responses to armed violence are couched in the context of a wider civil conflict. Violence there also operates in the milieu of large-scale international criminal organizations. In principle, this provides civic actors with relatively limited space to operate as they seek to reduce violence because of the scale of armed actors and how the state seeks to respond to those actors. That said, Colombia is also a robust democracy that offers civic groups space for engagement with like-minded actors in state and society as they seek to control violence. Medellín in particular has a strong history of civic mobilization in response to violence. Indeed, civil society has played an important role in bringing a series of mayors to power over the past fifteen years who have led substantial policy changes in the city. Still, the underlying civil violence has imposed some significant constraints on social mobilization both at the grass roots and in municipal-level civil society.
In Jamaica, gangs were largely formed by the two major political parties in efforts to control electoral districts through the armed distribution of patronage. These gangs remain closely tied to parties that also have links to factions within the police. The highly partisan nature of the Jamaican political system limits the number of strong civic groups that can work at the national and urban level to control violence. These dynamics lead to high levels of collusion between gangs and various state officials, which closes off Conditions: Significant occasional conflict between state and criminal actors amid some collusion and segmented sharing of governance; risk of inter-gang conflict; criminal groups exercise some control over civic groups.
Conditions: Very little public violence but significant targeted criminal violence; rare but explosive confrontations between state and illicit actors; criminals exercise significant control over policy and civic groups.
Strategies: Efforts build on collaboration between criminal and allied civic groups to target state violence; civic groups may also seek to work with armed actors to reduce their violent activities.
Strategies: Civic groups seek to collaborate with armed actors to control violence; when differences emerge between criminals and the state, civic groups collaborate with armed actors to protest the state.

Civic strategies under criminal disorder Civic strategies under tiered governance
Conditions: Significant inter-and intragang violence but little criminal control over civic groups Conditions: Conflict among divided gangs moderated by the state; civic groups largely independent but armed actor division complicates civic mobilization.
Strategies: Civic groups target gang behavior.
Strategies: Civic groups seek to work with gangs and the state to constrain gang violence. many pathways to controlling violence. This collusion and the highly partisan nature of politics substantially restrict space for civic mobilization. At the grassroots level, civic organizations are substantially dependent on criminal groups for access to the political system and for safety. At the level of wider urban civil society, partisan divisions substantially inhibit organizing and activism. Brazil has alternated between authoritarian and democratic regimes over the past half-century. Police have long been complicit in violence against the poor under both types of regimes, and courts have provided almost no relief to the poor from these abuses. Rather, some politicians have played important roles in working with the poor to protect their collective rights (Fischer 2008). In the current democratic regime, Brazil has seen the development of a relatively strong basket of rights for Brazil's poor amid substantial violence (Caldeira and Holston 1999). Police seek to arrest criminals but also engage in killings and abuse of the population. Rio has a robust history of civic mobilization around the city. At the municipal level, civil society is quite strong. At the grassroots level, despite a strong history of organizing, social groups are substantially constrained by violence. Indeed, hundreds of leaders of favela residents' associations in Rio have been murdered since the 1980s. Belo Horizonte has a more consensus-oriented political system but still has a relatively robust set of social organizations that are, on the whole, less constrained by violence that those in Rio.

Civic responses to violence
Civic leaders can choose between short-term confrontational strategies such as protests or more complex, long-term competitive or collaborative strategies to contain police or criminal violence. I will outline examples of six violence control strategies: protest against gangs, protest against police, complex confrontation with the gang, complex confrontation with state agents, constructive engagement with gangs, and constructive engagement with state agents. In the complex engagements described here, a number also involve civic actions that seek to control the root causes of violence through, for example, engagement with young people to give them opportunities that will induce them to avoid high-risk behaviors. I do not address these as independent violence-control actions since, in large part, they exist in the context of more explicit actions to control criminal behavior or constructively engage the state, and because their effects, when they operate independently of efforts to change violent actors, operate over a very long time frame and their aggregate effects are hard to interpret. Still I will highlight where these types of actions are undertaken as part of a broader violence control mobilization. While these strategies can operate independently, they are not necessarily mutually exclusive. I will explain the general conditions under which these actions arise. Table 2 outlines the six types of mobilization and their connection to different types of criminal governance structures.

Protests against police
Protests are perhaps the most common civic response to violence. Most frequently they are directed against police in reaction to the abuse, torture, or extrajudicial killings of individuals with little connection to crime. These actions are relatively easy to undertake and build on intense local concern. Police in democratic societies generally tolerate some degree of public protest, making this a viable short-term response. Finally, gangs usually support such protests since they are directed against police and because gang members and their families suffer much of this violence. Further, police violence can lead to backlash against gangs because inhabitants may blame the gang for attracting police violence. A typical example of antipolice protests occurred near the Cantagalo shantytown in Rio in 2001 after police shot and killed a young resident uninvolved in criminal activities. The community, as is the case with many favelas controlled by drug gangs, was experiencing divided governance and, consequently, chronic police-criminal confrontation. After the shooting, the police dragged the victim through the community. Residents came down the hill later to protest the murder and afterwards rioted, lighting a car on fire in a nearby middle-class neighborhood. Media reports gave prominent attention to the events. The state government was passing through a delicate political period as political allies, one of whom controlled the state government, were preparing to campaign against each other in the 2002 elections. As a result, the government changed policing strategy in the neighborhood, improving police-community relations over the medium term before things worsened again (Arias and Ungar 2009).
Antipolice protest also occurred in Tivoli Gardens in Kingston, Jamaica, in May 2010, when Prime Minister Bruce Golding declared his intention to use security forces to intervene in the area to arrest a gang leader politically connected to him that the United States had demanded be extradited. Collaborative governance, with close ties between state officials and the gang, prevailed in Tivoli Gardens. This collaborative relationship fell apart suddenly under intense international pressure. Residents protested the extradition effort and the impending police action (McGreal 2010). The protests failed and several days later the security forces intervened leading to the death of over seventy residents (Schwartz 2015).
Protests against police have mixed success. As the previous paragraphs show, public attention and successfully leveraging that attention over the government is critical to the success of these efforts. Ahead of an election, Rio's government was exposed to such pressure, whereas Jamaica's government, under countervailing pressure from the United States, was not.
The cases discussed above occurred, respectively, under conditions of divided governance and collaborative governance. Divided governance generates chronic conflict between security forces and armed actors and, as a result, yields various opportunities for protests against police. Generally, collaborative governance yields little conflict because of state-criminal alliances. However, at times when relationships change rapidly this can yield intense conflict as occurred in the Tivoli Gardens case. Police protest most distinctly emerges in cases where confrontations between police and well-organized armed actors put the population at risk. Civic leaders can lead these protests because these protests have gang support. The highly consolidated nature of gangs, however, forecloses space for protest against gangs. Protest is, of course, more successful if protesters have some sort of material leverage over the state such as the election campaign that began shortly after the protests in Cantagalo (on forms of competition of protest leverage see Eisinger 1973, 8-9).

Protests against gangs
Public protest is used much more rarely against gang violence. In 2014 residents of Denham Town and Tivoli Gardens, two impoverished neighborhoods in Kingston, Jamaica, protested against violence between competing gangs operating in their neighborhoods (Matthews 2014). Only four years earlier many residents of the same neighborhoods had turned out to protest police efforts to arrest Christopher "Dudus" Coke, a powerful gang leader who had dominated these areas for nearly fifteen years and who was held in high esteem by many in the local population for providing security and patronage (Campbell 2010). Coke's eventual arrest fractured local gangs and increased violence. The protest took place when gang structures had broken down and were generating violence. Despite these efforts, the gang conflict continues to rage despite significant police deployments (Ming 2016;The Gleaner 2016).
Rocinha, a large Rio favela several kilometers from Cantagalo, provides another example. In 2004 a conflict broke out between rival gang factions dividing the community. Scores died as a result of the conflict and confrontations with police. An important NGO at the city level worked with local grassroots groups to stage a protest in one violence-affected region of the community that involved a performance by a wellknown 1980s-era protest singer. Externally this event was largely billed as an event against the gang conflict. Internally, however, residents emphasized that the event was also a protest against police (Arias 2017). At the event the president of the local residents' association noted that the police are also "responsible for violence in the favela" and that police "abuse residents and invade their homes" (Monken 2004;Arias 2017).
Both of these cases show that antigang protests typically occur in contexts where gangs are weak and, as a result, in conflict. In particular, public protest against gangs is associated with criminal disorder or tiered governance in which gangs confront each other or police have little ability to control these confrontations. A single powerful gang that maintains order while it undertakes violence against enemies is unlikely to respond to protest, and protesters can become targets for retribution. Weaker gangs, however, are susceptible to protest. Where two similarly weak gangs confront each other, such violence can become chronic, undermine gang service provision, and cause gangs to prey on residents to accumulate resources (see Metelits 2010). This can generate protests that seek to shame gangs for their actions and pressure the government to adopt more effective containment strategies. At the same time, the Rocinha case shows why these types of protest are rare. Here, outside organizations whose leaders were not very exposed to gang violence billed the event as a protest against gang conflict, while the local association president emphasized the role of police violence in leading to the protest. Focusing on the role of the protest against police allows local leaders to nimbly protect themselves against potential gang retaliation since gang leaders, one of whom would eventually consolidate control over the neighborhood, are also critical of police violence. These efforts can lead to short-term cease-fires, but absent broader policy developments, such changes are unlikely to endure.

Complex strategies targeting gangs
Groups interested in controlling violence can also adopt more complex, long-term approaches focused on either gangs or police. Groups can constructively engage with the target group or constrain their behavior through competitive guidance. Constraining armed actors' behavior is challenging. If the competitive or collaborative efforts outlined here go wrong, gangs may target civic actors. Moreover, the process of undertaking these activities, regardless of success, is itself risky since gangs may undertake violence against individuals seeking to limit their operational space. Groups engaging in gang negotiations can also find themselves in legal jeopardy as a result of their own gang interactions.
Competitive approaches to gang constraint One strategy to control gangs is a multilevel alliance to constrain their activities. Grassroots organizations have the deepest understanding of gang behavior. These groups, however, often have the least capacity to promote action to contain gangs because of their limited social and cultural capital. As a result, these organizations need to quietly build alliances in broader civil society and with state officials to contain criminal groups. The connections that underlie these alliances enable civic groups to work behind the scenes with other actors to limit violent activities without directly confronting criminal groups.
Complex engagement against gangs has been an important strategy in Medellín, where civic groups have had to control paramilitary groups that had legitimately demobilized in a process that fostered armed group consolidation and strong ties between these groups and state officials, generating conditions of collaborative governance (Arias 2017). The government even promoted a participatory budgeting initiative to more effectively integrate marginalized communities into civic dialogues and governance. This process led to the involvement of gang members in councils allocating government resources (Moncada 2016b, 241-242;Abello Colak and Guarneros-Meza 2014, 3281). In some cases, demobilized paramilitaries sought funds to buy weapons, though in other cases they simply sought to control contracts through third parties so they could appropriate public funds. 1 Local leaders responded in various ways. Criminals expelled or murdered some of these community leaders. Others capitulated and either supported armed-actor-backed proposals or channeled funds to those actors from their contracts.
Medellín's residents developed varied responses. In Comuna Ocho, an area near downtown with a strong paramilitary presence, ex-paramilitaries such as John William "Memín" López exercised control over participatory budgeting, threatening civic leaders and controlling contracts. 2 Here paramilitaries took over community associations and, using local gang support, won elections to control the budgeting process. 3 Civic actors sought to contain their power by undermining their efforts to win control of the locality council and, as one civic leader noted, by using seats on the local council to disqualify paramilitary-tied groups 4 and by working with NGOs to increase state oversight of budgetary corruption. 5 These efforts helped the government identify corporations fronting criminal bids on participatory budgeting contract tenders.
Comuna Uno, on the city's northern frontier, faced similar problems with criminal groups threatening civic leaders to gain their support in budget negotiations and to give them government contract kickbacks. In many cases civic leaders lost control of their associations and were exiled. 6 A significant core of civic groups, however, sought to remain active by collaborating in multilevel coalitions to limit the armed actors' impacts on budgeting. Here a civic leader said some organizations had greater security against criminal threats. 7 A number of these organizations gained legitimacy in the communities where they operated by providing services to young people, including tutoring and sports classes, in the hopes of improving their life opportunities. By building ties between grassroots organizations and more powerful civic groups, these efforts enable some organizations to deflect armed groups' demands. These responses to the criminal manipulation of participatory budgeting in Medellín were effective at the moment, as local leaders sought to turn back illicit incursions. This was not, however, a comprehensive response to criminal interventions in this budgeting program. Criminal groups remain involved in Medellín politics at a variety of levels.
In the 2008 municipal elections Rocinha, which had long experienced divided governance with a wellorganized gang managing tense police relations, saw more limited impacts of a civic-political alliance seeking to undermine trafficker power. Here the gang leader backed an allied civic leader in a city council election and closed the community to competing electoral activity. This included numerous elected officials and several local leaders not aligned with the gang. Excluded candidates sought to highlight their concerns to state officials and the media. The police published documents reporting gang threats against residents who did not support the gang-chosen candidate (Ramalho 2010). A number of politicians, including a former major-party vice presidential candidate, appeared in the community to draw attention to the matter with police escort and press in tow (Tabak 2008;Mascarenhas 2008). These efforts, however, failed to dissuade the gang, and the gang's chosen candidate won office. One of the key critics of the trafficker position during this period left the community voluntarily after the election. Another remained in the community and continued to make efforts to advance his own dissident political career. Two years later the gang leader fabricated evidence that led to the arrest of this leader on weapons-trafficking charges, after he ran against the gang leader's chosen candidate in the state legislative election.
Competitive approaches to constraining gangs prevail when groups are well organized, as is the case under the collaborative and divided governance structures that existed respectively in Medellín and Rio in the above cases. The evidence presented here suggests that the success of civic containment efforts is driven by the degree of isolation of armed actors and the coherence of their civic opposition. In Medellín, participatory budgeting gave non-criminally aligned civic groups a basis for collaboration to derail armed actors' efforts to control the budgeting process, and to ensure, over time, that armed actors would have less ability to control government funds. The real dangers faced by activists in each community helped tie the anticrime network together and limit armed-group power. Conversely, in Rio the gang had a strong network of civic allies to support their criminal-political activities. Moreover, the civic-political opposition to gang activities was diffuse and self-interested, with politicians of different stripes using the election not to provide space for civic activists or enhance inhabitants' safety, but rather to draw attention to their own political projects. This suggests that the coherence of opposition and the ability of criminals to form alliances are key dynamics that can predict the success of these efforts. Further, these strategies are effective when civic groups engage with state actors that have the ability to control gang activities and interest in doing so, as occurred in Comuna Ocho, or in conditions under which they can build alliances with other civic groups operating in different political spaces, as occurred both in Comuna Ocho and Comuna Uno. Finally, these activities carry risks for the local leaders who collaborate against gangs, as was the case in Rio. To the extent that such activities are possible, local leaders need to feel that they have support and protection against potential future retaliation. In the cases discussed here, the two local leaders most clearly opposing the gang were well tied into alternative civic and political networks that provided them with some protection.

Constructive engagement with criminals
Civic groups can constructively engage criminals as they seek to channel their activities in more positive directions. This type of activity happens with surprising frequency though, due to its quasi-illicit nature, it receives little popular attention. Academics, however, have devoted some attention to this issue analyzing criminal truces (Durán-Martínez and Cruz 2016;Lessing 2015). Indeed, civic actors dealing with consolidated criminal groups frequently seek to engage with them. Efforts at collaboration have fewer short-term negative repercussions from armed groups than other types of efforts to constrain their activity, since such strategies acknowledge armed group power and build on that power to improve conditions. Since armed groups are embedded in particular communities, there are multiple pathways for civic actors to collaborate with them, though they have, of course, varying effects.
Medellín's Comuna Trece experienced a breakdown of criminal gang structures in 2009, leaving small gangs fighting one another amid criminal disorder. In this area, a civic leader, when asked about the challenge of producing an antiviolence concert during constant gang warfare, noted that he would negotiate with gangs to achieve a short-term peace so the concert could take place. The core of the movement here was oriented around local service and cultural organizations that provided outlets and opportunities for young people and the elderly. These efforts augmented the legitimacy of these organizations and gave them space in negotiating with violent actors. Civic leaders in another part of Comuna Trece negotiated with gang leaders to allow academics at the local university to carry out a survey. On a large scale, a group of civic leaders in Medellín, perhaps with the tacit support of government officials, initiated cease-fire discussions between crime factions in the lead-up to the locally hosted 2010 South American Games. Powerful politicians denounced these efforts despite the fact that they permitted the games to take place with minimal violence (Semana 2010).
Constructive engagement was similarly common in Kingston, where a history of gangs acting as semiformal political interlocutors facilitated this engagement. Here criminals controlled political patronage and were partially accountable to elected officials. A gang leader disobeying party orders might be arrested or killed. In Denham Town, which had a long experience of collaborative governance until May 2010, one civic leader said she always sought to engage the gang leader when she developed potentially controversial initiatives since you don't "want to go to war" without him "on your side." 8 In a divided community in eastern Kingston, a civic leader said that developing a consensus among gang leaders was important in starting new civic initiatives there. 9 In an east-side neighborhood with a highly consolidated gang leader with strong ties to the state, a civic leader built an alliance with a gang leader to undertake social programs in the gang's operational space. Finally, in 2002 the Jamaican government established the Peace Management Initiative in an area of Kingston affected by chronic gang conflict under conditions of tiered governance, where divided gangs had strong ties to state actors. This program negotiated truces among gangs in exchange for the government channeling development funds into their neighborhoods, some of which were devoted to programs to aid young people in violence-prone areas in the hope of providing them with increased opportunities, and to limit the numbers who would become involved in violent activities (Levy 2005). While such efforts were undertaken in various neighborhoods they were most successful along Mountain View Road, where the government intervened in a series of gang wars. Here a truce negotiated in a prominent hotel led to the formation of the Mountain View Development Council into which the government channeled development funds in exchange for an end to conflict (Levy 2005). 10 Local civic leaders worked with gang leaders and the state to implement these policies. These efforts achieved success as the intense gang wars that had rocked the area gradually declined.
Constructive engagement is relatively common, occurring in all four governance environments. When gangs fight in conditions of criminal disorder, engagement is often critical to calming tensions such as occurred in Medellín's Comuna Trece. This also took place with the divided warring gangs of eastern Kingston under the Peace Management Initiative. Finally, in cases where gangs are well organized, as in Denham Town, negotiating with their leaders is often one way of changing local conditions. Engagement with gangs is a particularly efficacious strategy for addressing violence. Gangs often have strong social roots in the areas where they operate. Individuals that confront gangs can face reprisals if they adopt confrontational strategies. Moreover, gangs often have a variety of interests, not all of which are at odds with local peace and stability. Most critically, gang members and their families live in the areas where they operate and suffer along with the entire community. Beyond this, maintaining some basic order is a pathway for gangs to establish security that will prevent residents from betraying them to police or other gangs. Constructive engagement recognizes these complex interests and adopts a realistic approach to bring armed actors into discussions about promoting stability in the communities where they operate.
Constructive engagement does, however, have some negative effects. Negotiations can legitimize armed actors, emphasizing their reasonableness rather than the violence they visit on an area. If their actions are contingent on government funds, as with the Peace Management Initiative, gangs may use those resources to strengthen their legitimacy and enrich themselves. Finally, civic actors that negotiate with gangs may lose some legitimacy by dialoguing with violent actors and, if they build relations with them, they may find themselves implicated in these groups' illicit activities. When peace breaks down, civic leaders may also be blamed for ineffective negotiations with armed actors. In this context, civic leaders need to consider how to undertake constructive engagement without further empowering gangs. This may involve insuring that funds destined to locales where these actors operate are channeled by civic groups, or ensuring armed groups take long-term actions to reduce violence. Finally, these efforts may also involve strategies to strengthen norms against violence.

Complex strategies targeting police
The last set of strategies to contain violent activities involves ensuring that police actively contain criminal violence in high-risk areas. As with interactions with gangs this can work through constraint or constructive engagement.

Constraining police
Similar to gang-constraint efforts, coercing police to improve their behavior involves cross-institutional collaboration. This includes working with civic groups that can control police behavior as well as working with actors and state-based networks that seek to promote police efficacy and control corruption.
In many cases, local organizations seek to confront police to reduce abuses or dissuade corruption. Due to the potential for retaliation by armed actors, activists often publicly confront rather than collaborate with police. Consistent and systematic actions against police, however, can also create challenges, making longterm confrontation with police hard to sustain.
Residents of Vigário Geral, a neighborhood in Rio de Janeiro that experienced divided governance as a result of conflict between a well-organized gang and the police, undertook such a strategy after the massacre of twenty-one community inhabitants by police in 1993. This had occurred shortly after gang members had allegedly murdered several police rather than paying a bribe. After the massacre, residents became active around Rio, attending meetings of the city's emerging antiviolence movement that developed in response to the increasing urban bloodshed after the 1985 return to civilian rule. Developing ties to other activists around Rio, residents undertook a forty-kilometer march along a highway from downtown to their outlying neighborhood to draw attention to the massacre. Residents also built ties with activists suffering similar problems with violence in the Acarí area about three miles away. These efforts enabled collaboration with outsiders to build a network including the Casa da Paz, a Médecins Sans Frontières office, and the Grupo Cultural Afro-Reggae's (GCAR) first favela-based center. A portion of these groups' activities, especially those of the GCAR, helped to focus on decreasing the exposure of young people to violence and reducing young peoples' involvement in illicit activities. These actions over the short to medium term helped increase the legitimacy of these groups in pressuring police and in making claims against criminals. These efforts helped residents promote, for a time, improved police responsiveness and a robust, though largely unsuccessful, prosecution of massacre perpetrators (Arias 2006;Ventura 1994).
While these efforts can have significant effects on police violence, they are particularly hard to bring to bear on police collaborating with criminal organizations, since such police-criminal alliances may target protestors. Indeed, the movement in Vigário Geral occurred precisely when police-criminal ties broke down. Working through networks to promote policy changes or legal action, however, can have significant effects and provide modest protection against retribution, especially when armed actors are weak, as was the case in Vigário Geral after the 1993 massacre.
Rio offers other examples of at least attempted complex constraining behavior. The Complexo do Alemão, on Rio's north side, has been the site of a remarkable amount of both police and criminal violence due to the area functioning as an important hub for the Comando Vermelho crime faction. On various occasions the police have violently intervened here. This violence attracted some investment from outside organizations that set up a small NGO network in the area, whose members activities focused on helping address some of the root causes of violence, including the problems of young people in the area. On one occasion I went to a meeting in the favela that brought together a variety of local leaders, sympathetic politicians, and city-level NGO representatives in response to a particularly egregious act of police violence that included the use of armored cars against gang-controlled positions in the community. The meeting ultimately did not succeed in gaining wider traction in organizing a systematic response to violence, due to the relative strength of criminal groups in the area and the police disrupting the event with another incursion. Similarly, the Acarí favela suffered a massacre in 1990 that led to the mobilization of a group of mothers whose children had been murdered by police, which became known as the Mães de Acarí. This group would develop relationships with various NGOs inside and outside of Brazil in their efforts to seek justice for their children and to constrain police behavior. Their organizing has become an enduring part of Brazil's growing movement against police violence.
These strategies are infrequent and often unsuccessful, as these cases demonstrate. In spaces of disorganized gang activity, where violence is the primary concern, civic groups may be interested in constraining police. Where the state and criminal groups collaborate, there is little space for complex police-constraint efforts. These actions are common under divided governance, such as existed in Vigário Geral, and in cases where relations between police and gangs are at an ebb, generating more open conflict and requiring civic organizations to promote residents' safety. The evidence presented here suggests that these events are most common in Brazil, where there exists a robust civil society well organized to constrain locally focused violent groups. In Kingston there is little in the way of independent civil society that could organize to constrain police behavior. To the extent that such actions were to occur, those demands would move through political patronage networks. Medellín has a robust civil society, but broader efforts to constrain state violence there are limited by the national security implications of police violence amid both a wider civil conflict and largescale international criminal organizations.

Collaboration with police
A final strategy is complex police collaboration. In this model, citizens and police collaborate through intermediaries to change policies to constrain armed activity. While these efforts are generally not secret, other actors tend to mediate them to build buffers between inhabitants and armed actors and corrupt police.
Brazil provides occasional evidence of this. The most notable emerged in Belo Horizonte's Fica Vivo homicide-control program. Here government intervention in one favela suffering criminal disorder driven by small-scale gang conflict established social programs to address the needs of high-risk youth populations and created a network of local leaders engaged with the state through recreational and educational activities. In short, the program sought to address root causes of violence in the context of a wider program that promoted civic-state collaboration. Largely this involved paying local youth leaders, some of whom had limited involvement in illegal activities, to run afternoon workshops for adolescents. These leaders attended training programs and developed a familiarity with local violence risks. At times, they reported these risks to social workers, who passed this information along to community police. The police then intervened to stop feuds from escalating (Arias and Ungar 2009, 424). Similar early-warning strategies also existed in Medellín.
Another example of collaboration with police emerged in Cantagalo after the protests against police in 2001 mentioned earlier in the article. Here the state government implemented a community policing that was accompanied by state investment in a variety of social programs. These efforts, which were based in a nearby state-run center, employed senior political leaders from the community to engage with area children in ludic activities and provided residents with better access to state services. Both Viva Rio, a major civic group, and eventually the Fundação Roberto Marinho, a philanthropic entity associated with the Globo television network, promoted networking between civic leaders and state officials and funded expanded social programs. The policing program at the time was led by a capable officer who engaged with local leaders and, indirectly, with criminals to reduce public violence. These complex engagements contributed, for several years, to substantially lower rates of violence (Arias and Ungar 2009).
These actions typically follow violent activities that undermine police legitimacy and expose them to public criticism. Thus, the issue here is not breaking police-criminal ties but rather networking with the government and ascendant reformist elements in the police to initiate proactive strategies to contain armed groups. Protests against police develop where police have been critical vectors of violence. Collaboration, on the other hand, emerges in spaces where, often after protests, there are empowered currents within police and among state officials that can collaborate with local leaders to promote efforts to control violence.

Analysis
The six categories of civic action discussed here highlight several factors in civic engagement that seeks to contain criminal activities and violence, both in the particular categories examined here and across the three countries. First, police are more often the subject of protest and constraining efforts rather than collaboration. While police often react negatively to protests, they do not live side-by-side with inhabitants and, perhaps more importantly, they are constrained, at least publicly, by legal norms. On that level, public protest or constraining activities highlight the obligations of police, thereby limiting retaliation. At the same time, direct collaboration with police to constrain gang violence could increase gang uneasiness, leading to retaliation against civic actors. All this adds up to police only being constructively engaged in efforts to control violence under unusual circumstances. The main example of constructive engagement with the police emerged in Belo Horizonte after the state government set up a series of local institutions to encourage and mediate community relations with police. In Rio it occurred amid pre-electoral politics that favored more progressive policies. Indeed, constructive engagement with police appears to be at least partially dependent on high-ranking officials initiating a new policy and seeking to reset community relations, thus opening space for dialogue. Given the history of police violence in these areas and ongoing criminal activities, these reinitiations of relations tend to be fragile, and future police errors or abuses have a tendency of making those relationships collapse again.
Second, an important characteristic of civic actions is that antagonistic efforts targeting gangs are uncommon. Gangs are only loosely bound by social norms prohibiting violence against peaceful civic activity. Moreover, these types of antagonistic civic engagements may expose criminals to attack by the state and by other criminals. The cases where these actions occur fall into two categories. For the most part, civic hostility toward criminals manifests itself when these groups are weak. This may occur, as was the case in Denham Town and Tivoli Gardens, when gangs are fighting with each other and residents intervene to demand that those gangs control their conflict. It may also occur when civic leaders establish a dense network of alliances both within their community and outside it to accumulate information needed to support officials' actions against criminal actors. Civic organizations may also operate in defensive solidarity that enables these groups to refuse criminal overtures and extortion but does not actually weaken armed actors.
A more common way to constrain violent activity is through gang collaboration. While broad social norms and the law place few limits on gangs, these groups are subject to local patronage and reciprocity norms. Inhabitants expect gangs to make some effort to minimize violence in the areas where they operate. Engagement with gangs plays on these local expectations. These collaborations thus enable grassroots leaders to communicate expectations in ways that can achieve gang compliance. The disadvantage of this approach, however, is that interaction with gangs generates legal and social liabilities for civic actors.
It is easier to engage in short-term protests than it is to undertake complex civic activities to change criminal or police behavior. These efforts often have little effect. Protests frequently occur when populations are upset about abuses, but, with important exceptions, short-term protests yield few long-term outcomes, given the underlying dynamics that produce criminal and police violence in the region. An important exception emerges when there are preexisting political dynamics that favor action. Where previous protests have taken place, where those protests occur in prominent locales, or, for one reason or another, officeholders become engaged with the protest, then short-term protests can produce longer-term outcomes. This was the case in the successful protests that led to a shift in policy that resulted in several years of improved policing in Cantagalo.
This discussion also reveals patterns across the three countries emerging from the history of statecivic engagement and the nature of violence in each locale. Protests of police occur frequently in all three countries. The costs of these actions are relatively low and can have some effect if undertaken at propitious times. Protests of gangs were rare because of the inherent risks of protesting armed actors operating in the same neighborhood where the protestors live. They were documented in one case in Kingston. In a case in Rio, protests occurred against gang violence, but local leaders emphasized that the protest was against police as well. Complex civic collaboration against gangs was common in Colombia, uncommon in Rio, and rare in Jamaica. This pattern is connected to local politics and this strategy's relative effectiveness. In Jamaica, gangs have strong political connections and, consequently, civic mobilization against them was futile and risky. In Colombia, on the other hand, local organizations had stronger bases for working with a relatively effective state against gangs amid a history of civil conflict and state action against organized crime and conflict actors. While the Colombian government also suffers corruption, with varied internal alliances and diverse patronage networks it is a more complex institution than Jamaica's smaller and highly partisan state. This environment enables some collaboration against gangs. Negotiations with gangs were common across the cases. In Kingston this occurred because gang-politician ties forced civic groups to negotiate with gangs to resolve various issues. Medellín also has a history of gang-civic group dialogues as a result of the history of conflict negotiations in Colombia. In Rio, gang-civic group negotiations are less common. Among the cases examined here, complex civic engagement against the state was a relatively rare strategy, emerging clearly only in Vigário Geral, though there are other examples in Rio. There was significant evidence of antiviolence collaboration with the state, particularly in Brazil. This arises from the central role the Brazilian state plays in social life.
Finally, the data reveal marked patterns associated with criminal governance. Protests of police were common where criminals were well organized. Protests of gangs emerged where criminals were weak and disconnected from the state. Efforts to engage constructively with gangs were surprisingly common and reflect the real power of these groups and the role of dialogue with these actors in controlling violence. Efforts to constrain gangs emerged where gangs were well organized and where civic groups constructed alliances to constrain their behavior. Cooperation with and complex constraint of police were relatively uncommon strategies as a result of the way these strategies could invite retribution by criminals. Constraint of police emerged only under divided governance where gang-police violence had become very high. Police engagement occurred under conditions of criminal disorder and divided governance where the police had gone out of their way to build community ties to help control violence. Table 3 shows patterns at the country and neighborhood levels.

Conclusion
This article has examined a variety of civic strategies to control crime-related violence. While the evidence does not show that any particular strategy is more effective than others, it does show that local conditions and broader political opportunities tend to constrain choices of the social efforts to control violence. Some of these dynamics are driven by the particularities of a country, city, or neighborhood.
The argument and data presented here offer a broad outline of civic responses to violence. Research into controlling violence requires more in-depth and systematic analysis of how civic groups approach criminal violence in the region. At the same time, the evidence I have presented suggests that the effects of these interventions are, at best, intermittent. Civic organizations can affect violence in a community but do not provide a substitute for state action. Indeed, over the long run, stresses on civic groups and changing conditions will cause many of these efforts to break down. Only systematic government and civic action offer a long-term violence-reduction strategy. Understanding the constructive relationship between the state and civil society in seeking to control violence and how state actors can facilitate civic responses to violence offers a potentially fruitful path for future research.