Determinants of Support for Extralegal Violence in Latin America and the Caribbean

What are the factors behind citizen support for the use of extralegal violence in Latin America? The prevailing argument is that, in countries overwhelmed by skyrocketing levels of criminal violence, people endorse the use of extralegal violence as a way to cope with insecurity. Other scholars believe that support for extralegal violence is the result of state withdrawal and failure. Few empirical studies, however, have tested any of these arguments. In this article, using regional data from the 2012 AmericasBarometer, we examine different explanations regarding citizen support for the utilization of extralegal violence in Latin America and the Caribbean. We developed a multi-item scale that gauges support for different forms of extralegal violence across the Americas, and we hypothesize that support for extralegal violence is higher not only in countries with extreme levels of violence but especially in countries in which people distrust the political system. Results indicate that support for extralegal violence is significantly higher in societies characterized by little support for the existing political system. Resumen ?Cuales son los factores que determinan el apoyo ciudadano al uso de la violencia extralegal en America Latina? El argumento predominante es que en paises con niveles altos de violencia criminal, las personas apoyan el uso de la violencia extralegal como un mecanismo para hacer frente a la inseguridad. Otros trabajos consideran que el apoyo a la violencia extralegal es el resultado del fracaso o ausencia del Estado. Pocos estudios empiricos, sin embargo, han tratado de probar dichos argumentos. En este articulo, utilizando datos regionales de la encuesta de 2012 del Barometro de las Americas, analizamos las diferentes explicaciones detras del apoyo ciudadano al uso de la violencia extralegal en America Latina y el Caribe. Desarrollamos una escala que contiene multiples items, la cual estima el apoyo a distintas formas de violencia ilegal en las Americas y planteamos la hipotesis de que el apoyo a la violencia extralegal no es solo alto en paises con niveles de violencia extrema, sino que lo es especialmente en aquellos en los cuales las personas no creen en el sistema politico. Los resultados obtenidos indican que el apoyo a la violencia extralegal es significativamente mas alto en sociedades caracterizadas por bajos niveles de apoyo al sistema politico existente.

What are the factors behind citizen support for the use of extralegal violence in Latin America? The prevailing argument is that, in countries overwhelmed by skyrocketing levels of criminal violence, people endorse the use of extralegal violence as a way to cope with insecurity. Other scholars believe that support for extralegal violence is the result of state withdrawal and failure. Few empirical studies, however, have tested any of these arguments. In this article, using regional data from the 2012 AmericasBarometer, we examine different explanations regarding citizen support for the utilization of extralegal violence in Latin America and the Caribbean. We developed a multi-item scale that gauges support for different forms of extralegal violence across the Americas, and we hypothesize that support for extralegal violence is higher not only in countries with extreme levels of violence but especially in countries in which people distrust the political system. Results indicate that support for extralegal violence is significantly higher in societies characterized by little support for the existing political system. ¿Cuáles son los factores que determinan el apoyo ciudadano al uso de la violencia extralegal en América Latina? El argumento predominante es que en países con niveles altos de violencia criminal, las personas apoyan el uso de la violencia extralegal como un mecanismo para hacer frente a la inseguridad. Otros trabajos consideran que el apoyo a la violencia extralegal es el resultado del fracaso o ausencia del Estado. Pocos estudios empíricos, sin embargo, han tratado de probar dichos argumentos. En este artículo, utilizando datos regionales de la encuesta de 2012 del Barómetro de las Américas, analizamos las diferentes explicaciones detrás del apoyo ciudadano al uso de la violencia extralegal en América Latina y el Caribe. Desarrollamos una escala que contiene múltiples ítems, la cual estima el apoyo a distintas formas de violencia ilegal en las Américas y planteamos la hipótesis de que el apoyo a la violencia extralegal no es solo alto en países con niveles de violencia extrema, sino que lo es especialmente en aquellos en los cuales las personas no creen en el sistema político. Los resultados obtenidos indican que el apoyo a la violencia extralegal es significativamente más alto en sociedades caracterizadas por bajos niveles de apoyo al sistema político existente.
Since 2010, in the southwestern Mexican state of Michoacán, thousands of people have joined vigilante groups or "self-defense forces" to protect their communities against criminal gangs and drug-trafficking organizations (Asfura-Heim and Espach 2013). During 2014, in the Argentinean city of Rosario and various neighborhoods of Buenos Aires, there have been more than a dozen cases of lynching of alleged criminals accused of robbery and drug trafficking (La Nación/Grupo de Diarios América 2014). These instances exemplify the extent to which violence exercised by nonstate actors has become a central feature of Latin America's current security crisis. Latin American countries have experienced an increase in violence and crime over the past thirty years, with homicides, robberies, and drug-related violence becoming a central concern for policy makers and citizens alike (Bergman 2006;Imbusch, Misse, and Carrión 2011).
In response, various governments across the region have decided to implement tougher anticrime measures. They have increased incarceration and police repression, have utilized army units to combat crime, and have allowed the discretional use of force by citizens and security forces against suspects. These harsh measures have proved ineffective in deterring crime or reducing violence. They have, instead, led to several human rights violations and the weakening of the rule of law (Dammert and Malone 2003;Hume 2007;Uildriks 2009;Valenzuela Aguilera 2013). Despite this, opinion polls demonstrate that draconian policies against crime enjoy citizens' general approval, as they are perceived as efficient and expeditious means to deliver security and justice (Dammert and Salazar 2009;Holland 2013;PNUD 2013).
Furthermore, citizens have increasingly incorporated private mechanisms of security provision to secure their material and physical integrity. These mechanisms include a broad range of responses, from private security guards and gated communities to vigilante organizations, self-defense forces, and the use of lynching against alleged criminals (Caldeira and Holston 1999;Godoy 2006;Goldstein 2003;Ungar 2007). The impact of these alternatives on crime levels has been mixed at best; at worst they have served to deepen already-unequal access to security provision. These options have also expanded the notion that criminals need to be stopped, even if it means bending the law (Adams 2012;Méndez 1999). Yet citizens remain supportive of these informal and often violent mechanisms of security provision.
What explains existing levels of support for extralegal violence in the region? Why, in some contexts, has the extralegal use of force come to be seen as a legitimate and efficient means to punish deviant-considered behavior? Mostly on the basis of qualitative research, several scholars have argued that support for extralegal violence is linked to increasing crime and violence across the region (Koonings and Kruijt 2004b;Ungar 2007). From a quantitative point of view, Bateson (2012, 583) has further demonstrated that in Latin America and the Caribbean, "recent victimization is strongly associated with rejection of democracy, support for authoritarianism, and approval of repressive policing and vigilantism." Other authors have pointed out that the key variable behind authoritarian responses is the perception of insecurity, not victimization per se (Ceobanu, Wood, and Ribeiro 2010;Malone 2013). In any case, the relationship between insecurity and extralegal responses may come as no surprise if we consider that crime has become more violent and visible across the region. According to the 2013 Human Development Report for Latin America, eleven of the region's eighteen countries have homicide levels considered at an epidemic level. Furthermore, robberies have tripled during the past decade, becoming the most recurrent crime that citizens experience (PNUD 2013).
Scholars have also advanced the notion that crime alone does not suffice as an explanation for attitudes that advocate the use of extralegal violence. Santamaría and Carey (2017), for instance, explain that perceptions and representations of crime across the region have contributed to legitimate and perpetuate both state and nonstate forms of violence and abuse. On the increase in cases of lynchings in different Latin American countries, Godoy (2004, 628) has argued that "lynchings are more a reaction to fear and insecurity than they are to crime per se." Meanwhile, for Arias and Goldstein (2010) vigilantism and human rights violations need to be understood in light of the economic, social, and institutional inequalities that have persisted in the region, which have contributed shaping the so-called violent democracies in Latin America. Last, several scholars (Bailey 2009;Cruz 2016;Dewey 2012;Waldmann 2006) have suggested that current levels of violence cannot be understood by looking only at criminal actors. They argue that it is important to analyze crime rates in regards to the impact that the state and political institutions have had in the production of violence in the region (Bailey 2008). For example, Zizumbo-Colunga (2015) has emphasized that citizens' distrust in law-enforcement institutions contributes to increasing citizens' support of vigilante justice in Mexico. In summary, scholars have pointed to three interrelated factors that might help explain the current approval for extralegal violence: insecurity and fear of crime, social inequality, and institutional weakness.
The goal of this article is twofold. First, we want to identify empirically some of the individual and contextual factors associated with people's support for the utilization of extralegal violence. Here, we understand extralegal violence as those violent and unlawful actions with the objective of punishing behavior that is considered deviant. Second, we test the prevailing arguments about the conditions that drive some people to support the use of illegal violence publicly. To do so, we use data from the 2012 AmericasBarometer, produced by Vanderbilt University's Latin American Public Opinion Project (LAPOP), and examine how crime victimization, trust in political institutions, and economic perceptions related to support for the utilization of extralegal violence in twenty-three countries of Latin America and the Caribbean.

Support of Extralegal Violence: Three Main Approaches
Scholars have pointed out three main explanations for citizens' support for extralegal violence in the region: fear of crime, social inequality produced by economic processes, and institutional weaknesses. Some of these explanations overlap in practice: people facing greater social and economic hardships may have less access to the justice system and, consequently, experience higher levels of fear or insecurity. However, we discuss these separately to highlight the different variables that can help us understand when and under which circumstances citizens' approval for extralegal violence emerges or intensifies. We use Latin America and the Caribbean as our laboratory to test these interpretations about endorsing violence.

Fear of crime
According to this approach, more than the actual levels of crime or violence affecting a country or a given locality, the levels of fear and insecurity experienced by citizens are what can help predict their approval and exercise of extralegal violence. Angelina Snodgrass Godoy (2004) offers this explanation when analyzing the recent increase in lynching violence in Latin America. In particular, she explains that in Guatemala lynchings are not concentrated in those areas with reportedly higher crime rates (i.e., they occur with more frequency in the western highlands than in the capital), and thus cannot be interpreted as a direct reaction to higher levels of crime. Furthermore, she argues that lynchings are organized in response to minor property crimes as opposed to those crimes that account for the higher levels of violence and crime in the country, such as organized crime. The same can be said of other scenarios such as Mexico, where lynchings are mainly organized as a response to petty robberies rather than to crimes linked to drug trafficking or kidnapping (Molzahn, Rios, and Shirk 2012;Rodríguez Guillén 2012). In other words, extralegal violence is driven by citizens' perceptions and subjective experiences, not only-or necessarily-by actual levels of crime.
In a similar line, Hume (2004Hume ( , 2007 has studied how local perceptions of crime in El Salvador tend to focus on the violence perpetrated by gang members, whereas people disregard other sources of insecurity such as intrafamily and intimate partner violence, even when the latter two constitute a greater cause of violent deaths in certain neighborhoods. Thus youth gangs have become the primary target for extralegal uses of violence in the country, manifested either in the police abusing their use of force or in targeted assassinations carried out by social cleansing groups and vigilante organizations. Arias and Marston (2017) identify a similar dynamic in Rio de Janeiro and Kingston, Jamaica. According to these authors, public perceptions of crime have centered on gangs operating in marginalized neighbors in both these cities, whereas other offenses, such as the use of death squads by political elites, have been largely ignored by the media. As explained by Dammert and Malone (2003) in the context of Chile, fear of crime is mediated by several social factors, including unemployment, poverty, trust in institutions, community participation, and social identity characteristics, all of which go beyond actual levels of insecurity. Krause (2014) likewise found that exposure to crime news in Guatemala influenced support for vigilantism by lowering citizen trust in government institutions.
According to Caldeira and Holston (1999), fear of crime may increase citizens' support for self-help justice such as lynching and vigilante groups, but it may also create incentives encouraging the use of extralegal violence by state actors. As they express it, this reaction may seem paradoxical, as "people usually want the police-whom they fear and accuse of being violent-to be violent 'toward the side that deserves it,' even though they know that the police routinely aggrieve innocent people" (Caldeira and Holston 1999, 712). This reality is also present in countries such as the Dominican Republic, where citizens consistently report high levels of distrust toward the police for their alleged corruption and arbitrariness, but nevertheless approve or tolerate bending the law to capture criminals (Bobea 2011). 1

Social inequality and economic distress
Another strand of the literature emphasizes economic factors behind citizens' approval of extralegal violence. It particularly focuses on the existing levels of social and economic inequality in Latin America as a primary driver of vigilante justice. In Bolivia, Goldstein (2004) has argued that, in a context where the provision of security has become privatized and defined according to class lines, lynching and self-help justice have come to be seen as a legitimate means to respond to crimes such as robbery, kidnapping, and rape. This is particularly true in those marginalized communities that have traditionally been excluded from the criminal justice system and that lack the resources to access alternative mechanisms of private security such as gated communities and private guards. Binford (1999) offers a similar explanation in the context of Mexico. He explains that lynching and support for vigilantism can be read as a direct consequence of the structural adjustments implemented in Mexico, which generated greater levels of economic exclusion and weakened the legitimacy of the security and justice institutions. Lynching and other more organized forms of vigilantism are thus interpreted as a reaction articulated by marginalized communities who respond to the social and political abandonment of the state by taking justice into their hands. As explained by Binford, lynching reproduces violence and distrust in communities that are already negatively affected by their social and economic vulnerability and has thus contributed to the further criminalization and policing of areas that are identified as ungovernable spaces. More recently, Phillips (2017) found empirical evidence that inequality spurs vigilantism, as economically disadvantaged populations feel relatively deprived of security compared to their wealthier neighbors who enjoy access to private security and even adequate public security.
The use of vigilantism and other illegal forms of self-help justice is certainly not exclusive to marginalized or impoverished communities, as political and economic elites have also exercised and authorized it. As several authors have asserted (Arias and Goldstein 2010;Benson, Fischer, and Thomas 2008;Costa 2011), in Latin America violence has served to perpetuate a given social and economic order as well as an instrument for overcoming the problems of exclusion that the system has generated. In light of this interpretation, elite groups' use of extralegal violence can be understood as an instrument to protect or secure their particular interests at a given time. Social cleansing groups-associations of armed individuals with the purpose of eliminating people considered "undesirable"-who operate in some Latin American countries have been linked to public officials and economic elites who identify certain forms of crime as an obstacle or as a destabilizing force that needs to be controlled and possibly eliminated (Huggins 1997;Waldmann 2006;Zilberg 2007). In the Mexican state of Michoacán, the surge in self-defense groups since 2012 has been attributed to large landowners and farmers whose economic interests are being affected by the extortions and kidnappings carried out by drug-trafficking organizations operating in the area and who thus decided to confront and expel these groups from their communities. In Brazil, the use of private security guards and police brutality to control crime has been analyzed in light of an urban upper class that is willing to earn more security even at the cost of human rights and democratic values (Caldeira 2001). In summary, according to this interpretation, social inequality is not only expressed in marginalized groups' approval or exercise of extralegal violence; it may also manifest itself in elites' willingness to bend the law and use illegal forms of punishment to secure their interest and economic privilege.

Institutional weaknesses
A third explanation of citizens' support for extralegal violence has focused on the weaknesses and challenges affecting Latin America's justice and security apparatus. Accordingly, this body of literature has highlighted the impact that high levels of impunity and corruption have had on the legitimacy and efficacy of these institutions and citizen approval of alternative mechanisms of security provision (Carreras 2013;Cruz 2011;Koonings and Kruijt 2004a;Ungar 2013). Ungar (2007), for instance, has argued that private mechanisms of security provision, including lynching but also private guards, gated communities, and so-called neighborhood patrols, are linked to Latin American states' failure to reform their security institutions. Among other things, this failure is related to the volatility and rotation characterizing security bureaucracies as well as with the pressure experienced by democratically elected governments that aim to reduce crime and violence during their time in office (Saín 2006;Ungar 2007).
For Davis (2006), democratization has had unintended consequences for the rule of law and has paradoxically contributed to the emergence and reproduction of self-help mechanisms of justice. In referring to the Mexican context, she argues that support for extralegal violence may be considered a reaction to the bureaucratic impasse, lack of coordination, and inefficacy generated by the decentralization and democratization of decision making in Mexico. Along the same lines, Pearce (2010) contends that democratization in Latin America has led to the authorization and legitimation of extralegal uses of violence, as governments build their authority on citizens' demands for more security and their "successful" battles and encounters with alleged criminal actors. Moreover, because democratic transitions reproduced the logic of social and economic exclusion that characterized Latin American's authoritarian past, people do view security not as a public good but as an instrument to exclude the many and protect the few. Whitehead (2009) adds that citizen's distrust in the state capacity or willingness to protect citizens has created the conditions for vigilantism and other forms of "rough" justice. Citizens regard extralegal uses of violence as a corrective to the state's incapacity to deliver justice and to exercise an effective form of punishment against alleged criminals. By the same token, Azaola (2004) explains that, in Mexico, citizens' distrust toward security officials, and particularly members of the police force, increases their willingness to bend the law to assert their sense of safety. Zizumbo-Colunga (2015) further expands this argument by explaining that support for vigilantism results from both citizen distrust toward state authorities and the presence of greater levels of interpersonal trust. It is important to highlight that the emphasis of this literature is not necessarily on the state's lack of presence in given areas but on the legitimacy of these public institutions in a given context. This may explain why in countries such as Bolivia, Guatemala, and Mexico there have been attacks against police officers or police stations by mobs trying to take justice into their own hands even when the police have apparently done their job by capturing a suspect.
This article intends to test these theories using Latin Americans' views about the use of different forms of extralegal violence. Most research referenced in the previous pages has focused on very valuable ethnographic work that, nonetheless, concentrates on specific countries. However, what about the overall region? Is it possible to extrapolate these theories to the hemisphere given that informal violence seems to have spread out to different subregional contexts and countries?

Data and Method
Our data source is the 2012 AmericasBarometer Democracy Survey, which taps perceptions and events regarding crime victimization, perceptions about security, opinions on institutions, and attitudes toward crime. The AmericasBarometer Survey is a regional project, unique in its scope, nature, and commitment to the scientific method. 2 The surveys from which the data are drawn were carried out in twenty-three Latin American and Caribbean countries during the first half of 2012. In each country, stratified probabilistic national samples were designed to represent the population precisely. On average, 1,500 adult citizens over the age of sixteen years were interviewed per country, for a total of 37,041 interviews. 3

Measuring support for extralegal violence
When referring to extralegal violence, especially in Latin America, most discussions have focused on vigilantism (Huggins 1991) or some forms of abuse of authority from state institutions (Caldeira and Holston 1999;Chevigny 1997;Menjívar and Rodríguez 2005). These are manifestations of violence that mostly take place in public spaces. In this article, we extend and operationalize the concept to forms of violence that also occur in interpersonal relations or private settings, such as the household, as we intend to measure the support to the illegitimate utilization of force in different circumstances, most of them in response to threats of crime, but not exclusively. Research in authoritarianism has shown high correlation between attitudes toward domestic violence and attitudes that approve the use of force against deviant or socially undesirable groups (Altemeyer 1996;Hetherington and Weiler 2009). In this article, we develop a measure of attitudes toward extralegal violence understanding that they are part of broader cognitive systems that interpret the way society functions and determine people's behavior.
To measure support for extralegal violence, we utilized a scale that was originally developed in 1997 by a group of researchers at the Pan American Health Organization for the first regional comparative study on attitudes and norms toward violence in the Americas. The project was called ACTIVA (Cruz 1999;Fournier et al. 1999). In the original version, the battery showed good levels of reliability, and it helped identify attitudinal systems that coexisted with high levels of violence (Cruz 1999;Orpinas 1999). We included and expanded the scale in the LAPOP's 2012 AmericasBarometer surveys across the region, as part of the UNDP's project to collect information for the Regional Human Development Report 2013-2014 (see PNUD 2013).
The scale comprises a battery of six items inquiring about the respondent's level of approval of the use of violence in different situations, ranging from physical methods of discipline toward children to police use of torture ( Table 1). The battery was included in the surveys of twenty-three countries in Central and South America, as well as in the Caribbean nations surveyed by LAPOP. For each item, interviewees were asked to respond any of the following alternatives: whether they approve the behavior of somebody else using violence, whether they do not approve but understand it, or whether they neither approve nor understand it. Rotated factor loadings for the six items formed two factors, with one loaded on five of the six items, and a Cronbach's alpha of 0.751, suggesting support for the creation of our index of support for extralegal violence from these five questions. 4 The reliability coefficient varied across country samples, from 0.642 in Haiti to 0.859 in Panama.
We constructed the index of support for extralegal violence transforming the responses of every item into scores from 0 to 100: responses of approval received a score of 100; no approval but understanding, a score of 50; and complete disapproval, a score of 0. We then averaged the results of the five items together. Average scores close to 100 represent favorable attitudes toward the utilization of extralegal violence; scores approaching 0 mean disapproving opinions.
The objective of this article is twofold. First, we want to identify empirically some of the individual and contextual factors associated with support for the utilization of extralegal violence. Second, we also aim to test the existing arguments that explain the support the use of illegal violence. Thus, operationally, our dependent variable, support for the use of extralegal violence, captures opinions of approval to the use of violence in different circumstances. Table 1 presents the average score for every item in the constructed index. In general, people in the region are more inclined to support the lethal use of violence against individuals accused of child rape, as well as to support the use of torture to obtain information about criminal organizations. In contrast, beating an unfaithful wife as well as carrying out social cleansing operations received the lowest responses of public support in the regional sample. Figure 1 presents the average score of the index of support for extralegal violence per country. The results show an interesting distribution. In this article, we do not claim that attitudes toward the use of extralegal violence are a factor driving the homicide rates in the region. However, the data show interesting parallels between our index of support for extralegal violence and the levels of homicidal violence per country for 2012 (PNUD 2013). Three of the five most violent nations in the Americas in 2012 (Honduras, El Salvador, and Guatemala) recorded the highest averages of support for extralegal violence, whereas four of the five least violent countries (Argentina, Chile, Costa Rica, and Uruguay) scored the lowest responses of support for violence in the region.
They also suggest a greater correspondence with the regional ranking of human development, although in the opposite direction (PNUD 2013), which may give some validity to the economic argument behind vigilantism. The five countries with the highest support for extralegal violence are also among the countries with the lowest Human Development Index in the Americas. Conversely, Argentina, Chile, Costa Rica, and Uruguay, the nations with the least support for extralegal violence, hold the top positions on the Human Development Index in Latin America and the Caribbean. Although the Human Development Index and our measure of support for illegal violence are negatively correlated only at a modest level (r = -0.2175), this bivariate correlation is higher than the positive correlation with homicide rates (r = 0.1320). 4 We excluded an item referring to the approval of a parent hitting his or her children. Table 1: Items of support for extralegal violence.

Regional average (0-100)
Suppose that a man hits his wife because she has been unfaithful with another man. Would you approve of the man hitting his wife, or would you not approve but understand, or would you neither approve nor understand?

25.0
Suppose that a person kills someone who has raped a son or daughter. Would you approve of killing him, or would you not approve but understand, or would you neither approve nor understand?

53.3
If a person frightens his community and someone kills him, would you approve of killing the person, or would you not approve but understand, or would you neither approve nor understand?

38.6
If a group of people begin to carry out social cleansing, that is, kill people that some consider undesirable, would you approve of them killing people considered undesirable, or would you not approve but understand, or would neither approve nor understand? 28.0 If the police torture a criminal to get information about a very dangerous organized crime group, would you approve of the police torturing the criminal, or would you not approve but understand, or would you neither approve nor understand?

Putting Theories to the Test
We test the variables associated with the support for extralegal violence roughly following the three theories already outlined: fear of crime, socioeconomic disadvantage, and institutional weakness. For each approach, we developed a mixed-effects model that includes individual and contextual factors. 5 Accordingly, our first model examines whether fear of crime increases support for extralegal violence. The second model concentrates on social, economic, and development variables that may be related to support for violence. The third model integrates a series of variables that account for the actual and perceived institutional capacity of the state to fulfill its functions. We ran a combined model with the significant variables from the three previous models to determine which factors were most important in their association with support for extralegal violence ( Table 2). 5 One point to consider is that the number of country cases is relatively small, at twenty-three, and hierarchical analysis recommends not exceeding ten cases per variable (Rabe-Hesketh and Skrondal 2008); therefore, we limited our use of contextual variables to no more than two for each model.

Do victimization and perceptions of insecurity augment attitudes toward the justification of violence?
In predicting attitudinal response, we included individual measures of victimization as collected by the AmericasBarometer, as well as perceptions of insecurity, and whether people see drug-trafficking organizations, youth gangs, or common crime as the main threat to their security. Details on question wording and index construction for all variables can be found in the appendix. We recoded all these variables in values running from 0 to 100 to standardize the interpretation. We also controlled for gender, age group, levels of education, area of residence (urban versus rural), and income measured as quintiles of family income. Because exposure to news can influence both fears of crime and attitudes toward the use of violence (Funk et al. 1999;Krause 2014), we added the variable of individual exposure to news media as control. In the fear model, we tapped the country context by adding homicide rates for the year 2011. Following the notion that fear of crime is conducive to approving attitudes toward violence (Briceño-León and Zubillaga 2002; Godoy 2004), we expect that people victimized and frightened by crime and people living in environments ridden with criminality will tend to support more the use of extralegal violence than in other contexts. Model 1 shows the results for the fear of crime argument. As expected, we can see that individuals who have been victimized and alarmed by crime are significantly more likely to exhibit approving opinions toward the use of extralegal violence. Results also show that people who consider youth gangs and drug-trafficking organizations main security threats tend to support the utilization of illegal violence. Likewise, the results of the fear-of-crime model show that, at the personal level, people who believe that the situation of security has worsened in the country tend to support violence more than individuals who think otherwise. All these relationships are statistically significant, and personal victimization has, in comparison, the largest effect of these variables on the probability to support extralegal behavior (coeff. = 0.043). 6 Results also indicate that being male, young, with no higher education, and low family income significantly increase the probability of expressing favorable attitudes toward the use of violence. 7 Also, exposure to news turned out to be significant, although with a negative sign, meaning that people less exposed to news media are more likely to favor violence. At the contextual level, homicide rates are positively and significantly related to support for violence. People who in 2012 lived in countries with reported higher murder rates are more likely to view favorably the use of extralegal violence than are people who live in societies with a lower homicide rate. These findings provide support for the argument that the increase in illegal acts of violence, such as social cleansing and lynching, is related to the moral panic generated by cumulative perceptions of crime. Nevertheless, the findings refute the thesis that perceptions are more important than actual events of crime when explaining support for vigilantism and extralegal violence. In fact, according to the data, both perceptions and actual victimization independently affect disposition to support violent acts. In addition, the model demonstrates that certain perceived threats (youth gangs and criminal organizations) are particularly critical in shaping attitudinal responses.

The model of economic distress
Socioeconomic explanations of the prevalence of extralegal violence contend that greater social exclusion and economic scarcities set the conditions for the use of extralegal violence against vulnerable and disadvantaged groups (Phillips 2017). Thus, people will support the use of extralegal violence in environments where they compete for resources and where inequality and exclusion mark social relations and access to security (Koonings and Kruijt 2007). To test this line of reasoning at the individual level, we pay attention to the dummy variables created to control by age, education, and income, then compare their stance on the use of extralegal violence. We do this expecting that groups that are usually in a position of social and economic disadvantage will score higher in our variable of support for violence than better-off citizens. In addition, using four items on economic conditions, we added two economy-related variables: one on perceptions of the current economic situation and the other on views of the economy in comparison with the previous year. At the contextual level, we incorporated indicators of per capita gross domestic product per capita and the Gini inequality index per country. 6 Because the second-level error term in the model correlates with the dependent variable and some individual variables, some estimates for individual-level variables might be biased. Hence, we reran all models as fixed-effects models. The results, presented in the appendix, show no significant changes in the estimates of the individual-level variables, indicating that all estimates are unbiased. 7 In the age variable, the group from sixteen to twenty-five years old was used as the reference category; in education, higher education level was used as reference; and in economic status, quintile 5 was the reference category.
The importance of economic factors on attitudes toward the use of violence can be seen in Model 2. People who perceived that the financial situation at the individual and national level is in bad shape were more likely to support extralegal violence than people who viewed the economy as positive. Likewise, respondents who perceived that the economy had worsened in the previous year (retrospective perception of the economy) tended to favor violence. Interestingly, the findings confirm that males are more likely to support extralegal violence than females. They also corroborate that extralegal violence is viewed less favorably as people age, and that Latin American citizens with a college education tend to endorse extralegal violence less frequently than the rest of the population. Also, respondents with less income (quintile 1) are more likely to favor violence than are those in quintiles 2-5. 8 At the contextual level, attitudes supporting the use of violence are more prevalent in the poorest countries of the region. Holding all other variables constant, there is a significant correlation between economic situation (measured in this case by per capita GDP) and support for extralegal violence. There is also a statistically significant relation with the Human Development Index (HDI) (not shown here), which points to the importance of countries' social development on the configuration of attitudes toward violence. However, the Gini index of income inequality, though having a positive sign, did not return any statistically significant relation with our attitudinal measure toward violence. In other words, our findings do not provide evidence that income inequality in countries is associated with citizen support for extralegal violence.

The institutional-weakness or failed-state model
Prevalent explanations about the rise of different forms of extralegal violence in the region revolve around the idea of institutional weakness and states' inability to respond to their citizens (Bergman and Whitehead 2009;Davis 2006;Fruhling 2003). Although these explanations intersect with those about insecurity and fear of crime, the former emphasize the institutional framework that not only allows alternative forms of violence to erupt but also encourages them. Following Easton's (1975) classification of political support, we included some independent variables that tap into diffuse and specific institutional legitimacy. First, at the individual level, we added Seligson's (2000) measure of people's diffuse support for the political system. This index comprises six different items in the AmericasBarometer that collect views on the overall institutions of the political system (see the appendix for a detailed description). Also, following Zizumbo-Colunga (2015), we created a variable that accounts for specific support for incumbent governments; this variable also comprises items that inquire about government performance in creating jobs, fighting poverty, and tackling crime. We expect that low marks on the scales of diffuse and specific political support are associated with high levels of support for violence, an indication of people resorting to private violence in response to the state's perceived ineffectiveness.
Additionally, anticipating that people may assess government's efficacy by looking at their public services, we added a variable that integrated people's satisfaction in government's services such as public schools, health care, and public works. Also, this variable may work as a proxy of the state's reach. Finally, at the individual level, we introduced a variable that evaluates the frequency with which people reach out to the different government agencies for help. We hypothesize that people closer to government institutions are less likely to approve the use of violence to solve conflicts.
At the country level, we tested several variables to evaluate the institutional thesis. First, we used the rate of police officers per population for each country, as we expect that more police presence may be related to less support for violent self-help actions. Second, we utilized two different indexes of countries' institutional development: the Fund for Peace's 2012 Fragile State Index and the Economist Intelligence Unit's 2012 index of democracy. We ran separate mixed models for each of the countries' institutional strength variables, but in Table 2 we show only the results of the Fragile State Index.
Model 3 shows that, at the contextual level, people living in nations with high scores of institutional fragility, as measured by the Fragile State Index, are more likely to express more favorable attitudes toward the use of extralegal violence. These results are consistent with other institutional measures. When we replaced the contextual variable of fragile states with the Economist's index of democracy, we found that people living in democratic nations tend to approve extralegal violence less often. The data suggest that what stimulates extralegal attitudes is living under political systems that are fragile and likely unable to meet expectations from the population. However, the findings do not indicate a statistically significant relationship between the degree of police reach, measured as the ratio of police per hundred thousand inhabitants, and support for extralegal violence, suggesting that such attitudes may not be merely the result of police absence or limited state reach.
In fact, at the individual level, the results give credit to the idea that poor system legitimacy is likely behind elevated levels of citizen support for illegal violence. As expected, diffuse support for the system is negatively related to approval for extralegal violence; that is, people who trust the basic institutions of the political system are less likely to endorse the use of extralegal violence. A similar relationship exists between specific support for government institutions (support for the government) and approval for extralegal violence, although in this case it is not statistically significant. However, data show that people who turn to authorities for help are more likely to support the extralegal use of violence. The latter is an unexpected finding. A potential explanation is that people who approve illegal violence may have previously reached out to authorities without obtaining satisfactory responses. As a result, they may have turned to support violence following their frustration with the government's reaction (Goldstein 2003). In any case, these findings overall follow some of the expectations outlined by the institutional perspectives on extralegal violence. They point to the fact that citizens supporting self-help justice and violence are more likely to be found in countries with weak institutional capacity and with little trust in the fundamental political institutions.

An integrated model explaining support for extralegal violence
So far, the results testing each of the theoretical frameworks have given credit to all of them. That is, we have found evidence that victimization, fear of crime, poor economic performance, low levels of political legitimacy, and institutionally struggling states tend to be associated with attitudes more strongly favoring extralegal violence. However, these interpretations as to why people support the use of extralegal violence may not rule one another out. More frequently than not, they may operate simultaneously. Therefore, we wanted to know which conditions are more critical in shaping attitudes conducive to individuals taking justice into their own hands and encouraging violence. Direct victimization? Trust in the political system? Or perceptions of economic conditions? We combine all variables in Model 4.
Results indicate that, holding all variables constant, support for the political system and victimization have the higher coefficients among the individual-level variables in the model. In the case of support for the system, ceteris paribus, a person with no faith in the system will score nearly six points higher on our index of support for extralegal violence than a person with absolute confidence in the system. Also, a citizen victimized by crime will score almost four points higher in our attitudinal scale toward violence than a citizen who has not been the victim of a crime. Interestingly, the effects of the perceptions about the economy decrease considerably (in comparison with previous models) when the variables of institutional trust and victimization are present. In other words, the grade of personal confidence in the political system and experiences of victimization seem to be the most important factors determining people's inclination to approve the use of violence. The results are similar at the contextual level: the introduction of the variables of homicide rates and fragile states reduces the significance of per capita GDP. People who support the use of extralegal violence are more likely to live in societies with elevated murder rates and fragile states, but not necessarily in countries with struggling economies.
In any case, the importance of low levels of legitimacy of the political system cannot be overstated. Although victimization and insecurity-related variables are also related to our dependent variable, the combined model indicates that distrust in the political institutions and living in a country with weak institutions are associated more with attitudes favoring extralegal violence than any other condition tested here. In all Latin American countries, the more people have faith in the political system, the less they are drawn to extralegal responses. Even in countries such as Costa Rica, Chile, Uruguay, and Argentina, where the propensity toward extralegal responses is comparatively lower, robust institutions intervene to reduce attitudes favoring violence even more (Figure 2). Several authors have stated this argument in some way (Chevigny 1997;Ungar 2007;Zizumbo-Colunga 2015). However, these results indicate that institutional legitimacy refers not only to the police and particular criminal justice institutions but, more importantly, to the general political institutions of the system. In other words, the problem of institutional legitimacy goes all the way up to the top of the political system. The inclination toward extralegal violence is not only about the weakness of the police and the courts; according to our results, it is primarily about the state.

Conclusions
This empirical cross-national test of the theories that explain citizen support for extralegal violence, the first of its nature, provides varying endorsements to explanations as to why some citizens support vigilante actions, social cleansing operations, and torture by the police, among other forms of extralegal violence. Future research could test empirically how these different variables interact and reinforce one another on the ground and at the local level. However, from the 2012 AmericasBarometer data, we can conclude that citizens who experience victimization and fear of crime are more likely to support extralegal violence. What is more, according to the results, favorable attitudes toward different violent actions may not only be driven by direct victimization and insecurity; they also seem driven by environments in which homicides and violence are rampant. Such a relation creates a policy conundrum, as countries with high levels of criminal violence beget attitudinal systems that end up breeding more illegitimate forms of violence, creating a vicious circle. It also supposes a challenge for measurement as environments with high levels of violence may be precisely the result of attitudinal systems that privilege the use of force. We have not been able to address such limitation here, but we expect to do so in future research.
The most significant finding of this project is that the institutional framework very likely plays a crucial role in setting the conditions for the emergence and maintenance of extralegal violence in most countries of Latin America and the Caribbean. Specifically, support for the diffuse institutions of the political system seems to be more important than any other variable, including victimization, insecurity, and trust in the efficacy of the government. These findings have important implications for understanding the perpetuation of violence in the region and for assembling policy responses to public insecurity. Much of institutional literature addressing criminal violence has focused on limitations of law enforcement and justice systems, as if the problem of violence were merely an issue of incompetent and corrupt organizations. The results shown here indicate that there is more to it. They suggest that the system's legitimacy may be a key factor explaining people's permissive views toward self-appointed vigilantes, armed groups, and violent responses to conflict. To the extent that states are unable to build legitimizing relations with their citizens, the lure of violence will continue.   Average predicted values support extralegal violence Support for the system To what extent do you respect the political institutions of (country)?
To what extent do you think that citizens' basic rights are well protected by the political system of (country)?
To what extent do you feel proud of living under the political system of (country)?
To what extent do you think that one should support the political system of (country)?

Support for the government
To what extent would you say the current administration fights poverty?
Average score scale 0-100 To what extent would you say the current administration promotes and protects democratic principles?
To what extent would you say the current administration combats government corruption?
To what extent would you say the current administration improves citizen safety?
To what extent would you say that the current administration is managing the economy well? (Cont.) Personal variables

Requested help from authorities
In order to solve your problems have you ever requested help or cooperation from…? [Read the options and mark the response] Average score scale 0-100 A member of Congress/Parliament A local public official or local government for example, a mayor, municipal council, councilman, provincial official, civil governor or governor) Any ministry or minister (federal), state agency or public agency or institution Have you sought assistance from or presented a request to any office, official or councilperson of the municipality within the past 12 months?
Satisfaction with state services In this city/area where you live, you are satisfied or dissatisfied with … The state of roads and highways Average score scale 0-100 The quality of schools The quality of the health system