1 
October 2021
This publication was produced at the request of the United States Agency for International Development. It 
was prepared independently by the Center on Conflict and Development, Texas A&M University. 
Vivian Bronsoler Nurko, Omar García-Ponce, Shahriar Kibriya, Isabel Laterzo, Elisa Lavore Fanjul.
Phase II Report 
Performance Evaluation of USAID/Mexico’s Crime and 
Violence Prevention Activity
Photo credit: JPV 
2
Contents
Abstract........................................................................................................................................................3
Executive Summary ................................................................................................................................4
I. Introduction...........................................................................................................................................9
II. Literature review............................................................................................................................ 11
II.1 Criminal Life Paths .............................................................................................................................11
II.2 Attribution of Blame..........................................................................................................................13
III. Context: Youth, Crime, and Violence in Urban Mexico............................................. 15
IV. Research Design............................................................................................................................. 17
IV.1 Research Questions and Hypotheses .........................................................................................17
IV.2 Research Design..................................................................................................................................18
Survey ................................................................................................................................................................................................... 18
Vignette Experiment................................................................................................................................................................... 19
Focus Groups.................................................................................................................................................................................... 19
V. Findings................................................................................................................................................ 22
V.I Survey Analysis .......................................................................................................................................22
Sociodemographic Characteristics of Respondents.............................................................................................. 22
Violence Normalization........................................................................................................................................................... 24
Protective Factors: Organizational Affiliation and Program Participation........................................ 28
Gang Participation...................................................................................................................................................................... 30
V.2 Experimental Results..........................................................................................................................33
V.3 Focus Groups ..........................................................................................................................................40
Violence Normalization........................................................................................................................................................... 43
Analysis of selected mediators ............................................................................................................................................ 47
Gang Participation...................................................................................................................................................................... 52
Blame Attribution........................................................................................................................................................................ 53
Protective Factors as Promoted by JPV Supported Interventions............................................................... 57
VI. Conclusions and Policy Implications.................................................................................... 61
VII. Recommendations....................................................................................................................... 63
References............................................................................................................................................... 66
Appendix................................................................................................................................................... 72
3
Abstract
Joining a criminal organization or participating in any kind of criminal activity involves numerous 
risks. Why do some young individuals decide to take these risks and embark on criminal life 
paths? A first step toward answering this question is to understand how youth rationalize violence 
and criminal behavior and under what conditions they justify involvement in crime. Relevant 
literature suggests that both contextual factors (e.g., socioeconomic condition, exposure to crime 
and violence during childhood) and individual incentives are important determinants of youth 
involvement in criminal life. Following standard theories of blame attribution, we use a mixed￾methods approach to investigate the motivations and justifications for youth violence and crime 
in urban Mexico.
First, we conduct a face-to-face survey representative of youth in urban Mexico—with 2,880 
individuals aged 16 to 29 in approximately 100 urban municipalities—to understand how 
socioeconomic characteristics and potential protective factors against delinquency relate to 
violence normalization and gang participation. Second, we embed a vignette experiment to test a 
set of hypotheses about blame attribution for criminal acts. Third, we conduct seven focus groups 
to further explore the mechanisms that mediate attitudes toward violence and drive the 
justification and blame attribution of crime involvement. We also explore how protective factors, 
as incorporated into prevention interventions, may mediate these attitudes. We explore which 
intervention components may work and why.
Our study contributes to a better understanding of patterns regarding Mexican youths’ exposure 
to violence and crime, how these individuals rationalize violence and involvement in criminal life, 
as well as how they attribute blame for different criminal acts. The study also underscores the 
role of normalization of extreme forms of violence, perceptions on the state and societal 
responsibility for youth involvement in criminal organizations, and the relevance of local crime 
dynamics in shaping these phenomena. These findings are crucial to design better policy responses 
for crime prevention. We generate actionable recommendations from our results.
4
Executive Summary 
The choice to participate in a criminal organization or engage in any kind of criminal activity 
involves numerous risks, including potentially life-threatening situations. This is particularly true 
for states with low governance capacity, poverty, and strong organized crime presence. Why do 
some young individuals decide to take these risks and embark on criminal life paths? A first step 
toward answering this question is to understand how youth rationalize violence and criminal 
behavior. If and to what extent do they normalize violence? Under what conditions do they justify 
involvement in crime? And how do they perceive the role of the government and society in youth 
crime? 
Following standard theories of blame attribution, we use a mixed-methods approach to 
investigate the motivations and justifications for youth crime in urban Mexico. First, we conduct 
a face-to-face survey of representative youth in urban Mexico—with 2,880 individuals aged 16 to 
29 in approximately 100 urban municipalities—to understand how socioeconomic characteristics 
and potential protective factors relate to violence normalization and gang participation. This 
observational data also provides representative information about trends with respect to youth 
exposure to violence, their relationships with people who might be involved in gangs, and where 
they are exposed to violence. Second, we embed a vignette experiment in the survey to test a 
set of hypotheses about blame attribution for criminal acts. Third, we conduct a series of seven
focus groups in Chihuahua, Ciudad Juárez, Guadalupe, Monterrey, and Mexico City to further 
explore the mechanisms that mediate attitudes toward violence and drive the justification and 
blame attribution of crime involvement. We also explore how protective factors against 
delinquency, as incorporated into prevention interventions may attenuate these attitudes. 
Further, our study explores the efficacy of different intervention components and the conditions 
needed for them to work.
Violence Normalization and Gang Participation
The results from the observational survey suggest that normalization of violence is more likely 
influenced by specific individual-level characteristics and experiences and less so by recent 
community-level violence. Individual-level characteristics such as gender, age, and socioeconomic 
background, as well as exposure to violence in one’s home appear more influential. Our results 
also suggest that young men, those with poor family relations, and those who have been exposed 
to domestic violence may be most at risk for violence normalization. While beyond the scope of 
this analysis, a more fine-grained data on individual exposure to different forms of community 
violence and domestic violence in follow up studies would improve the understanding of the 
potential mechanisms that influence violence normalization. We remain cautious to acknowledge 
that these cross-sectional findings based on correlations are not causal relationships. 
The focus groups results confirm and expand on these findings. Our analysis identifies four 
scenarios of violence normalization. The first is when violence is not normalized at all; that is, 
participants recognize it and actively reject it. Then, we categorize three types of violence 
normalization based on the extent of normalization, defined as nonrecognition, habit, and 
acceptance. The nonrecognition of violence happens when an offense or aggression is not perceived 
as violence but as some other form of social interaction. The habit of violence is when violence 
is recognized but is seen as a habit, or something that “everybody does.” The acceptance of 
violence is when violence is perceived as inconvenient but necessary to solve problems. These 
5
forms of normalization are mediated by the automated responses of youth, as well as by their 
beliefs, views, and values. 
We analyze in detail certain mediators of violence normalization that were identified as highly 
relevant, namely gender-based violence, domestic violence, substance abuse, and social norms. 
The focus groups disentangled a layer of violence normalization that is gender-based and rooted 
in gender stereotypes and misogynist views about how women and men should behave and their 
respective roles in society. This gender-based violence is prevalent in all social spheres where 
these youths participate—the community, work, school, and among friends, family, and in intimate 
partner relationships—which conflicts with the recognition of violence as such. As expected, 
gender-based violence is particularly intertwined with domestic violence, reflecting misogynist 
role models that reproduce gender stereotypes and drive violence. The evidence suggests that 
gender-based violence is learned at home, which, in turn, plays an important role in how 
normalization takes place. Expanding on the survey results, we find that previous offenders are 
highly exposed to violence in their families and that these violent dynamics are transmitted from 
one generation to the next. Violence is learned through socialization from family, and if someone 
grows up in an environment where violence prevails, it is likely that he or she will replicate it. 
We also find that substance abuse is seen as something that everybody does and, furthermore, 
that it is expected that someone who is high or drunk will engage in violence. Notably, substance 
abuse is also intertwined with family dynamics and plays a major role in generating violence among 
relatives. Group discussions showed that social norms in the social spheres of these youth 
reinforce beliefs and automated behavior associated with violence, particularly among peer 
relations at school and in neighborhoods. 
The survey results show that men typically know more gang members than their female 
counterparts, but that this is also influenced by the security of the community (the more insecure 
the community is perceived, the higher the percentage of youth who know gang members), and 
that those who know gang members consider joining gangs to be more fun. As socioeconomic 
status (SES) and educational attainment increase, so does the perception that gang participation 
will result in death and injury. The focus group analysis allowed us to further analyze the rationales 
behind youth’s perceptions of the disadvantages and advantages of joining gangs. We identify that 
economic benefits, coercion, and protection are at the core of their risk assessments and that 
these rationales vary depending on the criminal organizations’ dynamics presented in each 
context (compared to Mexico City, in the north, criminal organizations have monopolistic control 
over the territory and impose more coercion or potential to protect than in Mexico City). The 
results from the survey also point to the potential role of social networks as enablers of gang 
involvement and that this varies by type of organization. This finding may suggest that youth 
involvement in gangs does not happen in isolation, and it is plausible that certain local community 
organizations serve as hotbeds for gang recruitment. At the same time, it is also very likely that 
other types of organization or group activity disincentivize gang involvement. Based on the 
available information, we cannot disentangle the specific types of community organization that 
may facilitate or be correlated with normalization and gang involvement, but this is a relevant 
area of opportunity for future research. 
6
Blame attribution
The evidence from our experiment on blame attribution reveals that, on average, youth in urban 
Mexico see lower-class peers who engage in criminal behavior as less deserving of blame 
compared to middle-class, more privileged individuals. In such cases, respondents are more likely 
to blame the government or society as a whole for the crimes committed. This finding is also 
confirmed in the focus groups: while youth always assigned responsibility to the perpetrator for 
his choices, they tend to provide more justification if the perpetrator came from a lower 
socioeconomic background and acted out of need. Respondents are also more likely to blame 
perpetrators who act as gang leaders than those who follow orders. This result is consistent 
across all experimental analyses and in the focus groups. 
In the survey experiment we also find that the perpetrator’s actions are evaluated asymmetrically 
depending on whether the respondent is located in a low- vs. high-homicide area. While in high￾homicide areas respondents assign more blame to a middle-class perpetrator, this result is not 
significant in low-homicide areas. Similarly, individuals in areas of high violence perceive 
economically disadvantaged perpetrators as less to blame for their actions; rather, the 
government and society were seen as being at fault. The focus groups provided more nuance to 
this result. The views from the participants are congruent with the findings on normalization 
mediators, as society is seen to influence people’s behaviors. Interestingly, the concept of society 
in the focus group discussions was intertwined with family and personal relations. Youth blame 
the environment in which one develops. Moreover, youth blame the government for lack of 
opportunities and social services as leading youth offenders to make the choices they do. 
The survey experiment show that the type of crime and the identity of the victim seem to be less 
influential and highly context-specific in shaping assessments of attribution of blame. Namely, 
relationships were revealed when we examined participants divided by the safety of their 
community (these relationships were obscured within the pooled sample). Blame assigned to the 
perpetrator increases with crime severity in low-homicide areas but decreases in high-homicide 
areas. The focus groups also provided nuance to this result. While most youth express more 
blame with increasing crime severity, many of them reduce this blame considering that the 
perpetrator likely did not have a choice. Youth justify that when one belongs to a criminal 
organization, there is no other choice but to follow the leader’s orders, also showing a sense of 
vulnerability. The focus groups also provided more insight into how the type of victim shapes 
crime perceptions. This happens mainly through two mechanisms: empathy and risk assessment. 
Youth show empathy toward more vulnerable victims (an elderly lady and a blue-collar worker) 
because of their social class reflecting their own backgrounds. In this sense, attacking a 
businessman or a politician would entail a reduced moral load. At the same time, youth expressed 
that a businessman and a politician are riskier victims because they are more likely to have 
bodyguards and personal connections in the government to enact punishment for the crime.
Further research is needed to explore whether these highly contextual differences with respect 
to the victim and the type of crime are because of increased normalization of violence or because 
of a generalized stigmatization of all criminal behavior. Taken together, these findings show that 
youth rationalize criminal behavior as the result of structural conditions and not only individual 
agency. The focus group discussions also revealed the complexity underlying the rationalization 
processes of these youth. While most youth recognize the centrality of individual choices, society 
and the government were also blamed for crime and violence. This suggests that effective youth 
7
crime prevention requires trust-building strategies to restore the social fabric and legitimize state 
institutions.
Key Takeaways
Zooming in on how prevention programs may positively impact at-risk youth (particularly 
offenders), through the focus groups, we find that prevention programs are able to promote 
protective factors through the components of mental health services, addiction treatment, 
support networks, and education and employability. Nonetheless, several challenges remain. In 
terms of mental health services, we find that, while the current mental health services which are 
mainly focused on socioemotional health are effective, psychiatric treatment and work with 
families remain urgent. As mentioned above, domestic violence, substance abuse, and criminal 
activity are highly intertwined with family relations in this context. Although CSOs have 
incorporated family therapy in their service offerings, this is insufficient to address these complex 
family dynamics. Most of the youths we surveyed or interviewed had abused drugs and had been 
exposed to highly violent situations, either as perpetrators or victims. However, there was no 
indication of services addressing these problems present in their communities. Furthermore, 
addiction treatment remains a challenge and a sensitive issue to address, given the close 
relationship of substance abuse with violence and criminal behavior, particularly considering that 
it affects the social spheres of most of these youths. Harm reduction interventions seem to be
promising, although they are more the exception as opposed to the norm. Finally, in terms of 
employment, the main area of opportunity relates to work training and labor intermediation, as 
youth struggle to find or maintain a job. 
Along with these areas of opportunity, our study contributes to a better understanding of how 
youth in Mexico rationalize an individual's involvement in criminal life and how they attribute 
blame for different criminal acts. It also underscores the role of normalization of extreme forms 
of violence and the state and societal responsibility for youth involvement in criminal 
organizations, as well as the importance of local crime dynamics. Such findings are also crucial to 
design better policy responses for crime prevention and should be taken into consideration.
Based on our findings, we provide a set of recommendations. First, we encourage the 
international aid community to invest in efforts to map local crime dynamics and potential 
enablers for recruitment, crime, and violence. It is important to map out gangs at the 
neighborhood level, and to consolidate a census of neighborhood community organizations to 
identify correlations with gang activity in high-violence areas. These efforts should be advised by 
academic experts in these subjects and led by local actors. Second, it is imperative to rapidly 
expand the offerings of youth prevention programs, addressing the mechanisms that drive 
violence normalization. In this sense, multidimensional therapy models, addiction treatment, and 
psychiatric models adapted for the Mexican context are promising. We also strongly recommend
prioritizing gender-based violence prevention and augmenting interventions aimed at supporting 
victims of gender-based violence. Third, we recommend implementing perspective-taking 
interventions to inform participants about the drivers of youth involvement in criminal behavior 
and raise awareness about the risks and the impacts—for them and for victims—associated with 
such behavior. Examples include police-youth and victim-youth dialogue programs. Although 
these strategies may not be enough as stand-alone interventions for youth who have already been 
involved in crime, they may be helpful as a secondary prevention intervention. Fourth, we 
advocate for more private sector involvement to expand employment opportunities for at-risk 
8
youth, creating paid job-training programs and positions that are aligned with the labor market 
demand and train youths to fit those needs, improving profiling and reducing rotation. Fifth, we 
consider it urgent to engage with government agencies to create and expand specialized capacity 
for the implementation of specific services aimed at preventing youth crime involvement and 
reducing recidivism in higher-risk communities. Particularly, expanding services for mental health, 
employment, and cultural services is paramount, as well as creating programs that allow the 
combination of post-release support or alternatives to incarceration with comprehensive social 
services. Finally, we propose the creation of a collaborative research model with academic 
researchers and practitioners to improve the accumulation of data and knowledge to inform 
evidence-based policy interventions. 
9
I. Introduction
This report is the second installment of the performance evaluation of the United States Agency 
for International Development’s Juntos para la Prevención de la Violencia (USAID/JPV) aimed at 
promoting reductions in crime and violence levels by strengthening local prevention systems 
(LPSs) and reducing the likelihood of youth participation in crime and violence. The findings from 
the first phase suggest that crime in Mexico is a highly contextual behavioral phenomenon, and 
successful prevention interventions need to consider the broader context of violence to which 
program participants continue to be exposed during and after the program. Building on those 
findings and following Sampson and Laun (2005), we conceptualize crime as “a socially emergent 
and contextually shaped property … [a]n emergent process reducible neither to the individual 
nor the environment.” Within this perspective, “neither agency nor structural location can by 
itself explain the life course of crime,” and it is necessary to study them together. In this report 
we examine how socioeconomic and contextual variables mediate the normalization of violence, 
and following theories of blame attribution (McGraw, 1991), we explore how youth explain the 
causes of crime and violent behavior. We also explore how prevention interventions are able to 
promote protective factors and how these mediate attitudes of violence. Building on the findings 
of the first phase of the performance evaluation, we identify effective components of JPV 
supported interventions as well as areas of opportunity.
Criminal organizations often target young individuals from disadvantaged socioeconomic 
backgrounds as potential recruits. While some youths are recruited by force or other means of 
coercion, many others voluntarily decide to participate in criminal activity despite the risks of 
violence and arrest. A large body of research has been dedicated to identifying the determinants 
of juvenile crime, particularly in the United States (US) and other developed areas of the world. 
This literature suggests that both contextual factors (e.g., socioeconomic conditions and 
childhood exposure to crime and violence) and individual incentives are important predictors of 
youth criminal involvement. Several studies have provided evidence that material incentives 
influence crime engagement (Becker, 1968; Ehrlichh, 1973; Draca and Machin, 2015). Others have 
emphasized the role of nonmaterial motivations, such as status and respect in society and social 
and peer networks (Billings et al., 2019; Bruce, 2007; Cook and Ludwing, 2005; Lindquist and 
Zenou, 2019). The existing work is invaluable to understand aggregate patterns of youth crime. 
However, there has been limited research aimed at exploring both individual incentives and the 
rationalization of youth violence and criminal involvement. A first step to understanding why 
some young individuals choose this lifestyle is to examine how youth normalize violence and 
rationalize criminal behavior and under what conditions they justify it. This study contributes to 
filling this gap by investigating the motivations and justifications for youth crime and violence in 
urban Mexico.
In particular, this study aims to understand criminal and violent behavior in Mexican youth 
through the following inquiries: Why do some young individuals decide to take these risks and 
embark on criminal life paths? To what extent do they normalize violence? Under what conditions 
do they justify involvement in crime? How do they perceive the role of the government and 
society in youth crime? Are there differences by gender? This study also examines how 
interventions targeted at at-risk-youth can help strengthen and improve protective factors against 
crime delinquency.
10
To answer our questions, we use a mixed-methods research design based on three key 
components: (i) an in-person survey representative of youth in urban Mexico; (ii) a vignette 
experiment to test a set of hypotheses about blame attribution for criminal acts; and (iii) a series 
of focus groups to further explore the mechanisms that drive violence normalization and the 
justification and blame attribution of crime involvement.
The results suggest that normalization of violence is more likely influenced by specific individual￾level characteristics and experiences and not so much by recent community-level violence. 
Individual-level characteristics such as gender, age, and socioeconomic background, as well as 
exposure to violence in one’s home, seem to matter more. In general, young men, those with 
poor family relations, and those who have been exposed to domestic violence may be most at 
risk for violence normalization. We also find that gender-based violence, domestic violence, 
substance abuse, and social norms are important mediators that lead to violence normalization. 
Our experimental results show that, on average, youth in urban Mexico see lower-class peers 
who engage in criminal behavior as less deserving of blame compared to middle-class, more 
privileged individuals. In such cases, youth are more likely to blame the government or society as 
a whole for crimes committed. Youth are also more likely to blame perpetrators who act as gang 
leaders than those who follow orders. Finally, we find that the type of crime and victim identity 
seem to be less influential and highly context-specific in shaping assessments of blame attribution. 
Our results suggest that local crime dynamics influence these assessments and that these 
rationales vary depending on the criminal organizations’ dynamics presented in each context. 
In terms of protective factors as promoted by JPV-supported prevention interventions, through 
the focus groups we find that the components of education and employability, mental health 
services, addiction treatment, and support networks are effective at promoting protective factors.
Nonetheless, several challenges remain. In terms of employment, the main area of opportunity 
relates to work training and labor intermediation, as youth struggle to find or maintain a job. In 
terms of mental health services, we find that psychiatric treatment and work with families remain 
urgent. Addiction treatment remains a challenge and a sensitive issue to address, given the close 
relationship of substance abuse with violence and criminal behavior and considering that it affects 
the social spheres of most of these youth. Together with the findings above, our findings indicate
that prevention strategies should be tailored to the specific local crime dynamics, identifying 
venues of socialization that minimize youth risks and that are insulated from criminal groups.
The remainder of the report is structured as follows. In the next section, we review the literature 
on criminal life paths and attributions of blame, and in the following section, we review the 
context of youth, crime, and violence in urban Mexico. Section IV presents the research design. 
Section V presents the findings, and Section VI presents the main conclusions. Finally, Section VII 
presents a set of recommendations for future prevention programming. 
11
II. Literature review
II.1 Criminal Life Paths
Since the seminal contributions by Becker (1968) and Ehrlich (1973), a growing body of literature
has demonstrated that economic incentives are important determinants of crime outcomes. 
Income-based explanations underscore the role of wages and unemployment rates in the formal 
labor market relative to the economic returns and labor opportunities in illegal sectors (Draca 
and Machin, 2015). This literature has also provided compelling evidence of “neighborhood” 
effects, in which crime is highly concentrated in certain population groups and places (Weisburd 
2015)—particularly in poor neighborhoods with marginalized populations (Pettite and Western, 
2004; Raphael and Sills, 2007). For Latin America, Jaitman and Ajzenman (2016) find that in five 
cities of the region, 50% of crimes are concentrated in 3% to 7.5% of street segments. Calderon 
et al. (2018) confirm this finding for Mexico, with evidence suggesting that the share of homicide 
cases found in the top-10 most violent municipalities in Mexico rose from 20% in 2016 to nearly 
27% in 2017—representing the highest proportion of homicide cases concentrated among 
centers of violence since 2012 (when this figure was greater than 30%). Research has also shown 
that social interaction is a key channel through which neighborhood crime is linked to individual 
criminal behavior, where the share of convicted criminals affects later crime conviction 
probabilities, as well as the number of crimes for which a young man is convicted (Damm and 
Dustnamm, 2014). Similarly, several studies have shown that income inequality is associated with 
higher crime incidence (Freedman and Owens, 2014; Levitt and Lochner, 2001).
Some literature has focused on understanding what links crime rates with neighborhood 
disadvantage (Ludwig et al., 2001) and how crime involvement is affected by the proportion of 
criminals living in the same residential neighborhood (Damm and Dustmann 2014; Bernasco et 
al. 2017). This literature has pointed toward peer interaction models or social cohesion as 
variables that explain variation in crime rates among otherwise similarly disadvantaged 
neighborhoods. For instance, Glaeser et al. (1996) suggest that crime is reinforced by agent-to￾agent meetings that create information flows or interactions that result from input provided by 
family members and peers, which determine the cost and taste for crime. Sampson et al. (1997) 
argue that social and organizational characteristics of neighborhoods explain variations in crime 
rates that are not solely attributable to the aggregated demographic characteristics of individuals. 
They propose that the differential ability of neighborhoods to realize the common values of 
residents and maintain effective social controls is a major source of neighborhood variation in 
violence (p. 918).
The literature has further explored peer effects. Evidence has shown that the “source of crime” 
is located in the intimate social networks of individuals (Lindquist and Zenou, 2019). Peer effects 
can happen through friends, family members, or neighbors. There has been a growing body of 
literature showing that juvenile delinquency is a group activity (Lindquist and Zenou, 2019; 
Haynie, 2001; Warr, 2002). As Lindquist and Zenou (2019) explain, “delinquents often have 
friends who have themselves committed several offences, and social ties among delinquents are 
seen as means whereby individuals exert an influence over one another to commit crimes” (p.
78).
12
The school has been identified as one relevant space where these peer interactions take place. 
Existing evidence suggests that higher educational attainment is negatively correlated with crime 
involvement (Lochner and Moretti, 2004). The bulk of the literature has indicated that 
disadvantaged and at-risk students in the US tend to attend schools with higher dropout rates 
and lower academic achievement (Rothstein, 2004), which is consistent with findings from extant 
studies on Mexico (Cárdenas, 2011). The literature has also found consistent evidence of the 
peer effects happening at school, where the concentration of high-risk youth increases the 
aggregate level of misbehavior (Cook and Ludwing, 2005; Carrel and Hoekstra, 2008; Daming 
2011). Billings et al. (2019) show that social interactions explain the large variance in criminal 
activity across neighborhoods and find evidence for an agglomeration effect, by which grouping 
more disadvantaged students together in the same school in small neighborhood areas increases 
total crime. Daming (2011) find that better schools reduce future crime and presented suggestive 
evidence that the effect happens through higher school quality in high school and through peer 
effects in middle school. Furthermore, there is evidence that exposure to criminal violence has a 
negative effect on Mexican students’ academic achievement (Caudillo and Torche, 2014; 
Michaelsen and Salardi, 2020).
Peer interactions have also been shown to lead to other behaviors, such as alcohol and drug 
consumption, which is also related to crime involvement. Evidence has shown that there are 
contagious effects in which the probability that a young person behaves in a certain manner 
depends positively on the prevalence of such behavior among their peers (Becker, 1998, Crane, 
1991). Gaviria and Raphael (2001) find evidence of peer-group effects at the school for these 
behaviors, as well as school dropout. There is ample evidence showing that a high percentage of 
youth entering the justice system are drug users (McBride et al., 1999) and that drug use is also 
related to recidivism (Mitchell et al. 2007). In a study of gang members in Rio de Janeiro’s favelas, 
Carvalho and Soares (2015) find that individuals from lower socioeconomic backgrounds and 
with no religious affiliation have a higher probability of joining a gang, while those with problems 
at school and early use of drugs join gangs at younger ages. For the case of Mexico, Chávez 
Villegas (2020) find that gang members serving prison sentences show more materially inclined 
motives and higher substance consumptions relative to peers with no criminal record. Descriptive 
evidence has also shown that youth in conflict with the law present high rates of substance abuse 
have been exposed to domestic violence, and present high levels of school dropout (Azaola, 2016; 
CIDAC, 2016).
Other types of social interaction derive from family, and there is important evidence of an 
“intergenerational transmission of crime.” Parental human capital and behavior influence 
intergenerational crime relationships (Hjalmarsson and Lindquist, 2012 & 2013), and there is 
evidence of sibling correlations in crime involvement (Eriksson et al., 2016). The literature 
concerned with family effects on crime has found that adverse family influences driven by 
inadequate parenting and unstable homes affect crime involvement (Levit and Lochner, 2001, 
Blumstein, 1986). This relates to findings from studies on childhood risk factors, including parental 
substance abuse, that show how these are associated with future antisocial and criminal behavior 
(Stevens, 2018). Finally, evidence also has suggested that exposing children to illegal labor markets 
makes them more likely to be criminals as adults (Sviatschi, 2017).
13
II.2 Attribution of Blame 
The literature on the attribution of responsibility in political psychology has paid particular 
attention to citizens' attributions of the government's responsibility for economic performance 
(e.g., Arceneaux, 2003; Kluegel and Smith, 1986; Levin et al., 2016; Peffley, 1984). A growing 
number of studies have also explored blame attribution for government performance in 
noneconomic domains such as natural disaster management (e.g., Achen and Bartels, 2002; 
Malhotra and Kuo, 2008). Following McGraw (1990, 1991), another stream of research has 
examined judgments of contextualized actions of individual actors. For instance, Gibson and 
Gouws (1999) investigate the willingness of ordinary citizens to lay blame on a specific individual 
for violent actions during the period of struggle over apartheid in South Africa. Our study builds 
on these streams of literature, with a focus on the factors that affect individual-level blame 
attribution, as well as how blame may be shared across various actors. Our focus is not only on 
attribution of blame for the perpetrator, but also for societal actors, including the government.
Previous work in social psychology has examined the relationship between some personality 
variables and attribution of blame by offenders for their criminal activity (Gudjonsson, 1989), as 
well as the role of cognitive distortions (Blumenthal et al., 1999), type of offense (Felson and 
Palmore, 2018; Gudjonsson and Bownes, 1991; Bastian, Denson, and Haslam; 2013), and 
emotions (Lerner et al., 1998) in shaping attributions of responsibility. In these works, researchers 
have suggested that the severity of the crime can affect blame attribution; when examining blame 
attributed to the victim of a crime, less blame has been attributed for more severe crimes (e.g., 
rape vs. robbery) (Felson and Palmore, 2018). Further, convicted criminals seem to attribute 
more blame to themselves when they have committed more severe crimes (Gudjonsson & 
Bownes, 1991). Research on reactions to youth crime and perceptions of accountability has 
reinforced such results, finding that the type and outcome of a crime are major motivating factors 
in sentencing decisions and perceptions of legal competence (Ghetti and Redlich, 2001). The 
specific psychological mechanisms that enable youth to engage in acts of violence remain unclear, 
but recent studies have suggested that normalizing violence and dehumanizing victims partially 
explain the violent behavior of youth among gang members (Alleyne et al., 2014). 
Knowledge of the influence of contextual factors on blame attribution, as well as the various 
connections among human capital, family upbringing, and economic circumstances, also suggest
that the economic context of crimes may influence blame attribution. Some studies have 
suggested that individuals are particularly unforgiving when considering poor citizens, often 
blaming them for their own poverty; however, this depends on respondents' socioeconomic 
backgrounds (Cozzarelli et al., 2001). Also, recent studies have suggested that blame is often 
shifted to the government in contexts of economic hardship (Sirotkina and Zavadskaya, 2020).
Accountability mechanisms also seem to be of importance and can include assessments of 
causality, responsibility, and culpability (Shaver 2012). The role of the actor (whether they are a 
leader or are taking orders) clearly influences blame attribution, as there is a difference between 
what a person did and what a person was obliged to do (Hamilton and Hagiwara 1992). Delegating 
decisions, for example, can shift attributions of blame (Bartling & Fischbacher 2012). 
Finally, perceptions of the victim also influence blame attribution. In contexts of sexual abuse, 
research examining victim blaming has found that victim age, attractiveness, and history of abuse 
affect blame attribution (Rogers et al., 2007). Such factors also can affect attribution of blame to 
14
the perpetrator (Rogers et al., 2011). Within this general framework, more blame is expected to 
be attributed when victims of crime are perceived to be innocent (García-Ponce, Young, and 
Zeitzoff, Forthcoming). This research suggests that victim identity may also influence blame 
attribution for the perpetrator vs. society as well. 
15
III. Context: Youth, Crime, and Violence in Urban Mexico
Over the past decade and a half, various regions of Mexico have been heavily affected by a wave 
of crime and violence that affects the life choices of youth. Our understanding of this phenomenon 
remains limited, to a large extent because of the significant heterogeneity across the country in 
terms of criminal activity and socioeconomic structural conditions, which influence youth 
behavior and decision making. As presented in the previous section, the available evidence shows 
that both the environment and social interactions affect crime involvement through different 
channels, and emerging evidence in Mexico suggests that crime is correlated with the risk factors
known in the literature. Notably, gang violence encompasses much more than the commonly 
studied drug-trafficking organizations (DTOs) and organized crime groups (OCGs). While DTOs 
and OCGs are prominent, smaller gangs are often not connected, or are loosely connected, to 
such groups and engage in significantly less-high-profile crimes that result in lesser economic gain. 
Youth account for an overwhelmingly high percentage of perpetrators (and victims) of gang 
violence, with recent estimates in Mexico being that approximately 30,000 children and 
adolescents are actively participating in criminal organizations (IACHR, 2020). In addition, youth 
have been known to create their own gangs—often termed “youth gangs”—which typically seek 
to create spaces for their members to exercise rights and access opportunities that they do not 
find provided to them by their family, community, or government; in these efforts, they often 
violate the rights of others and create violence (Jones, 2013, OAS, 2007).
Gang structures vary by region in Mexico. Southern Mexican states such as Chiapas and Oaxaca 
have a significant presence of Central American gangs, while Northern Mexican gangs are heavily 
influenced by US gangs. Central Mexican gangs tend to be characterized as “youth groups” often
with minimal criminal activity. Northern Mexican gangs are structurally modeled on U.S. street 
gangs, but not every gang in northern Mexico has a strong connection to U.S. gangs nor are they 
as dangerous. The scant literature on youth gang involvement in Mexico suggests that 
socioeconomic and psychological factors drive youth gang involvement. Unemployment, limited 
access to the education system, lack of parental involvement, and poverty are associated with a 
higher likelihood of criminal behavior among children and adolescents (Jones, 2013).
The Mexican government’s response to the gang phenomenon has given priority to a law 
enforcement approach—“social policy” proposals have been carried out primarily by 
nongovernmental organizations (NGOs). Of note, President Felipe Calederón (2006–2012) 
declared the “war on drugs” early in his administration, dedicating significant resources to military 
deployment to combat organized crime and gangs. In sharp contrast with his predecessor, Enrique 
Peña Nieto developed the National Plan for Crime Prevention, placing the prevention approach 
at the center of the security strategy and assigning resources to municipalities to develop social 
prevention strategies.1 However the security strategy of both the federal and state governments 
1 The National Program for the Social Prevention of Violence and Crime (PRONAPRED) subsidized social prevention 
policies at the municipality level. Between 2012 and 2016, the program distributed more than 10 billion pesos to 
local governments all over the country, but the budget sharply decreased over the years to almost disappear by the 
end of the presidential term (Ayala & López, 2016). Evaluations of this program suggest that its efficacy was very 
limited given a poor technical design (Chapa Koloffon and Ley, 2015).
16
during this period was more of the same. More military operations, more army, more security 
forces developed at the state level without training or evaluation (Le Cour, 2018). President 
Andrés Manuel López Obrador followed the same approach and similarly relied on law 
enforcement, deploying a new national guard to combat violence. Across administrations, such 
strategies have targeted not only large DTOs and OCGs, but have impacted smaller, local gangs 
as well. Critics have pointed out that this strategy ignores the root causes of crime and violence 
and often breaks up larger organizations and breeds smaller gangs that compete for territory and 
resources. Such groups often engage in crime that increasingly affects the local communities, such 
as participating in the retail drug market and engaging in extortion.
With regard to social policy strategies, the Mexican government’s response has been limited in 
comparison to repressive approaches. The government has dedicated resources to programs 
aimed at reducing inequality, a factor that has been shown to contribute to increased violence in 
Mexico (Enamorado et al., 2014). For example, the Prospera (previously Oportunidades) program 
targeted families in extreme poverty and aimed to decrease inequality and increase school 
attendance and healthcare visits of children. Ultimately, the program did result in important 
outcomes, such as increasing the average time in school of young males by 10 months 
(WorldBank 2014). However, despite these efforts, Mexico’s performance in this domain, 
comparatively speaking, is lacking—with continuing high rates of inequality and poverty, low rates 
of educational attainment, and issues of social and healthcare inclusion (SGI 2021). Furthermore, 
programs and social policies specifically targeted at violence prevention among vulnerable 
populations are limited and are often provided by the NGO sector.
17
IV. Research Design
IV.1 Research Questions and Hypotheses
The goal of the study is to understand how youth rationalize their peers’ involvement in criminal 
life in order to be able to design tailored, impactful interventions. We are particularly interested 
in gaining a better understanding of the following questions: To what extent do youth normalize 
violence? Under what conditions do they justify involvement in crime? And how do they perceive 
the role of the government and society in youth crime? 
The outcome of interest can be conceptualized as attribution of responsibility or justification of 
engagement in crime. Justifications are important for blame attributions and are highly contextual 
(McGraw, 1991). Within the context of criminal behavior, as previously discussed, background 
characteristics such as SES help in understanding youth involvement in the criminal world. The 
type of involvement or the role of the actor (whether they are a leader or are simply taking 
orders) also influences blame attribution, as there is a difference between what a person did and 
what a person was obliged to do (Hamilton and Hagiwara, 1992). Within this general framework, 
more blame is expected to be attributed when victims of crime are perceived to be innocent 
(García-Ponce, Young, and Zeitzoff, 2020) and when the crimes committed are more severe or 
outrageous (Bastian, Denson, and Haslam, 2013).
Building on these theoretical foundations, we generated the following testable hypotheses about 
individual responsibility:
Hypothesis 1. An individual from a disadvantaged or marginalized socioeconomic 
background will be judged as less responsible for the crimes they committed, relative to 
an individual from a middle-class or affluent socioeconomic background. 
Hypothesis 2. An individual following orders from a criminal gang leader will be judged 
as less responsible for the crimes they committed, relative to an individual with a 
leadership role within the criminal gang.
Hypothesis 3. An individual who perpetrated crimes against victims who are perceived 
as innocent will be judged as more responsible for the crimes they committed, relative to 
an individual who perpetrated crimes against victims who are not perceived as innocent.
Hypothesis 4. An individual who perpetrated more severe crimes will be judged as 
more responsible for the crimes they committed, relative to an individual who
perpetrated crimes that are not as severe.
Following similar logic, we also tested the following hypotheses about societal and governmental
responsibility:
Hypothesis 5. If a crime were committed by an individual from a disadvantaged or 
marginalized socioeconomic background, compared to a middle-class socioeconomic 
18
background, the government and society will be judged as more to blame relative to the 
individual.
Hypothesis 6. An individual following orders from a criminal gang leader, compared to 
an individual with a leadership role within the criminal gang, will be judged as less 
responsible for the crimes they committed relative to the government and society.
Hypothesis 7. If a crime were committed against victims who are perceived as more 
innocent, the government and society will be judged as less to blame relative to the 
individual.
Hypothesis 8. If an individual commits more severe crimes, the government and 
society will be judged as less to blame relative to the individual.
IV.2 Research Design
Our research design is based on three key components. First, we conduct an in-person survey 
of representative youth in urban Mexico to measure and analyze their sociodemographic 
characteristics, victimization, attitudes toward violence, in/security perceptions, exposure to 
violence (at home and school), exposure to and perceptions about gang participation, and 
protective factors against delinquency as promoted by prevention programs. Second, within this 
survey, we embed a vignette experiment to test our hypotheses about blame attribution for 
criminal acts. Third, we rely on a series of focus groups following a sequential explanatory design 
to further explore the mechanisms that drive the justification and blame attribution of crime 
involvement. These focus groups target beneficiaries from JPV-supported interventions.
Survey 
We survey a population that is representative at the urban national level for ages 16 to 29 (n = 
2,800). Our sampling design ensures that our sample is not only representative across common 
sociodemographic categories (e.g., education and income), but also by level of violence. To do 
so, we consider three variables that capture levels of violence at the municipal level: homicide 
rate, reported nonhomicidal crime, and perceived level of violence. Homicide rates are 
considered more accurate official statistics compared to nonhomicidal crimes, as they are often 
reported more often by the general population and are typically recorded more accurately 
because they are definitionally specific and typically go through the health system (UNODC 
2019). However, this measure does not capture the full reality of insecurity.
For this reason, we also include measures generated from Mexico’s National Survey of Urban 
Public Security (ENSU) to capture nonhomicidal violence and insecurity at the municipal level. 
Given that the ENSU data are not representative at the municipal level, using this survey and the 
2015 intercensus, we generate municipal estimates using multilevel regression and 
poststratification (MRP). These measures capture the preponderance of nonhomicidal crime 
(MRP victimization) and perceived community insecurity (MRP insecurity) at the municipal level. 
With these estimates and homicide rates, we then order municipalities based on level of 
insecurity and sample via seriation. Our sampling strategy generated a survey sample that is 
reflective of the ENSU survey in terms of violence level across all three categories. Section A in 
19
the Appendix presents more detailed information about the data and methodology used for the 
survey design.
As mentioned above, the survey is composed of four blocks that aimed to gather information 
about respondents’ sociodemographic characteristics, victimization, attitudes toward violence 
(violence normalization), in/security perceptions, protective factors against delinquency, 
exposure to violence (at home and school), and exposure to and perceptions about gang 
participation. The questionnaire (in Spanish) is available in Section B in the Appendix.
Vignette Experiment
To test our hypotheses, we embed an original vignette experiment in the survey. Experimental 
vignette studies in survey research use short descriptions of hypothetical scenarios (vignettes) 
that are usually presented to respondents within surveys in order to elicit how their judgments 
about such scenarios affect outcomes of interest, often revealing their perceptions, values, or 
social norms. In our vignette, we randomize the perpetrator’s SES/upbringing, the type of criminal 
involvement (leader vs. gang member), the severity of the crime, and the type of victim to 
understand how youth attribute blame. The following table presents the vignette with randomized 
segments in bold. In total, we had 48 experimental combinations:
Ahora te voy a contar una historia de una situación que pasa frecuentemente en muchos lugares de nuestro país. Cuando 
termine de leer, te pediré tu opinión sobre la situación.
Rodrigo es un joven mexicano que creció en una familia [DE BAJOS RECURSOS CON MUCHAS CARENCIAS/DE CLASE 
MEDIA DONDE NO LE FALTÓ NADA].
Desde muy temprana edad empezó a delinquir y recientemente fue arrestado por [ROBARLE EL CELULAR / EXTORSIONAR 
/ SECUESTRAR / ASESINAR] A [UN OBRERO/ UN EMPRESARIO / UN POLÍTICO LOCAL].
Rodrigo confesó a las autoridades haber cometido este crimen [Y SER EL LÍDER DE UNA BANDA CRIMINAL / Y DIJO 
QUE ESTABA SIGUIENDO ÓRDENES DEL LÍDER DE UNA BANDA CRIMINAL].
We then ask respondents three questions to assess the influence of each attribute on the 
assignation of blame to the perpetrator (Rodrigo), the government, and society using these 
variables. We isolate the main effect of each variable across the pooled sample. In addition, we 
isolate the main effects for certain subgroups of interest according to other variables we gathered
in the survey or to our municipalities' strata. In particular, we compare these effects for more 
violent vs. less violent municipalities and men vs. women. Power analyses indicate that our sample 
size is large enough to find effects for all main effects (controlling for interactions) across both 
the pooled sample and smaller subgroups (if divided into two groups). In particular, a power 
analysis (at the 0.8 level) indicates that an n of 1,397 responses is needed to find results at the p 
< 0.05 level. With a sample of 2,800, we have a sufficient sample size to conduct analyzes, even 
when dividing our sample in half.
Focus Groups
Following a sequential explanatory design, we conduct six focus groups with beneficiaries of JPV￾supported interventions and one additional focus group with youth involved in one of Mexico 
City’s Puntos de Innovación, Libertad, Arte y Educación (PILARES) groups. The target population 
20
of the focus groups is young offenders (as these were youth participating in one tertiary 
prevention program given their trajectory) and from low socioeconomic backgrounds, which 
means that this analysis should be interpreted as a zoom-in on this segment. These participants 
were contacted through civil society organizations (CSOs) that either intervene at the community 
level in highly insecure areas, that provide services for alternative justice measures to detention 
for young offenders, or that do both types of intervention through a vast array of services and 
activities, such as therapy, education, scholarships, labor intermediation, sports, and cultural 
activities, among others. The PILARES group is an exception. PILARES is a policy of Mexico City’s 
Department of Education that integrates 25 community centers located in prioritized areas in 
terms of insecurity. These centers offer services such as distance learning, work training, 
conversation groups, remediation courses, and cultural activities. Participants from the PILARES
group are at-risk youth who live in low-SES neighborhoods in Mexico City, but who have not 
been involved with criminal activity and who have more protective factors than youth from other 
groups. We include this group to be able to gather some data on youth from low-SES
neighborhoods who have not yet been involved in crime. The neighborhoods of all participants
are insecure, and fights between neighbors and drug dealing are part of daily scenarios.
The groups were conducted after the survey was completed to further explore the mechanisms 
of violence normalization and blame attribution. The focus groups also explore how protective 
factors, as promoted by crime and violence prevention interventions, mediate these attitudes. In 
interviewing direct beneficiaries, we aim to understand which intervention components might 
have worked (or not) and why in order to decrease the likelihood of youth engaging in violence. 
Focus groups convened both women and men, with the number of participants ranging between 
4 and 11 per group. Although the design intended to have a gender balance, this was only achieved 
in one group; the rest had a majority of men, as shown in Table 1. 
Table 1. Focus group participants per program
Organization Program description Women Men Total
Centro de Asesoría y 
Promoción Juvenil, A.C. 
(CASA)
Alternative measures 
programs/community 
intervention
3 7 10
Consejo Ciudadano 
Seguridad y Justicia A.C.
Alternative measures 
programs
3 4 7
La Tenda di Cristo A.C. Alternative measures 
programs/community 
intervention
2 4 6
Puntos de Innovación, 
Libertad, Arte y Educación 
(PILARES)
Community centers 4 4 8
Reinserta a un Mexicano 
A.C.
Alternative measures 
programs
1 6 7
Renace - Solidaridad y 
Justicia
Alternative measures 
programs
1 3 4
Supera A.C. Alternative measures 
programs/community 
intervention
2 5 7
21
The focus group sessions discussed youth hobbies, work, goals, satisfaction with CSO programs, 
and violence normalization in different settings. Also, groups were presented with a vignette to 
assess blame attribution similar to that which was present in our survey. For the violence 
normalization section, participants were divided by gender into two separate groups, as there 
were questions related to gender-based and sexual violence that might have generated biased 
responses in mixed groups. The vignette was adapted from the survey, and it was presented as 
an evolving story introducing variation in the variables of interest as the discussion progressed, 
namely in the perpetrator’s (Rodrigo’s) socioeconomic background, the type of victim, and the 
severity of the crime and its circumstances (e.g., following orders, making his own decisions). The 
focus group guide (in Spanish) is available in Section C of the Appendix.
22
V. Findings
V.I Survey Analysis
In this section, we examine the sociodemographic characteristics of respondents and their 
relationship with violence normalization, as well as the relationship of protective factors 
(understood as participation in prevention programs and organizational affiliation) with violence 
normalization. We also explore how youth risk assessment of gang participation relates to their 
exposure to gangs (risk factors) and to protective factors. 
Sociodemographic Characteristics of Respondents
In this section, we explore the socioeconomic characteristics of our respondents, particularly 
examining their work status and gender. For this analysis, SES is measured via a series of questions 
regarding respondents’ possession of certain household goods (e.g., television, cell phone, vehicle) 
and was created using principal component analysis (PCA). Overall, we find that fewer than half 
of the sample was employed during the fielding of the survey, with some variation by gender and 
SES. 
Figure 1 presents the work status of respondents broken up by gender. This figure 
demonstrates the plurality of both men and women having worked in the past week (across both 
genders, 43.5% of respondents worked). A minority of respondents neither worked nor were in 
school (6.3% across both genders). This figure also shows women as disproportionately dedicated 
to housework (across both genders, 15.6% of responders were dedicated to housework). 
Figure 2 also presents the work status of respondents via a box-and-whisker plot and breaks 
up respondents by SES. The middle horizontal line of each bar represents the median SES of each 
group. From this figure, we can see that students tend to be of higher SES. Among women, those 
who are dedicated to housework tend to be of the lowest SES, while among men, those who are 
incapacitated and unable to work represent the lowest SES. Figure 3 presents the employment 
type of individuals who reported having an occupation in the past week. The majority of both 
males and females were employees during this timeframe. About one-quarter of respondents 
were self-employed. Overall, from these figures, we can conclude that the main difference 
between males and females is the overrepresentation of females in housework and that this is 
more pronounced in lower-SES conditions. Moreover, youth from higher-SES conditions tend to 
stay in school.
23
Figure 1. Work status of respondents by gender
Figure 2. Work status of respondents by gender and SES
24
Figure 3. Employment type of individuals who reported having an occupation in the past 
week
Violence Normalization
This section explores violence normalization. We first present average perceived levels of 
violence by act and by gender. Then, we show a series of regressions examining how 
normalization is influenced by victimization (robbery at home or business, robbery of car, assault 
on public transportation, being hurt by a weapon, attempted homicide, extortion, kidnapping, 
sexual assault), perceived level of insecurity, school victimization, domestic violence, and 
interpersonal trust—all of these controlling for a set of socioeconomic characteristics. We find 
that violence is significantly more normalized among men compared to women, with some 
variation dependent on the type of crime examined. We also find some evidence that abuse at
the home (verbal or physical) is associated with violence normalization, as is school victimization. 
Ultimately, trust in one’s family is associated with less normalization of violence (i.e., those who 
trust their family more normalize violence less). These results suggest that young men, those with 
poor family relations, and those who have been victims of abuse at home may be most at risk for 
violence normalization. 
Figure 4 presents average perceived levels of violence by act (nonconsensual sex, hitting on 
someone, hitting someone, nonconsensual kissing, selling drugs, and making fun of someone) and 
by gender. Nonconsensual sex, hitting, and selling drugs are considered the most violent acts 
across genders, whereas hitting on, nonconsensual kissing, and making fun of other people are,
on average, considered less violent. However, it is noticeable that, in general, women tend to 
consider all acts more violent than men do.
25
Figure 4. Normalization of violence (by act, gender, and age)
Table 2 further examines normalization of violence based on victimization, perceived level of 
insecurity in the community, and sociodemographic controls, including SES, age, gender, and 
respondent education. In addition to the acts presented in Figure 4, this table includes greeting 
26
or helping someone one doesn’t know. In the survey, we include these variables to get more 
variation in the responses and to avoid bias if all items were negative. The regression in this table 
and all regressions in this section employ ordinal logistic regression models (utilizing the polr 
package in R). All models include fixed effects by state. The variable “victim” is a binary variable 
that equals 1 if respondents reported any type of victimization in the past 12 months and that 
equals 0 otherwise. The community security scale is from 1 to 7, where 1 is very insecure and 7 
is very secure. The normalization scale is from 1 to 7, where 1 is not violent at all and 7 is violent.
To our surprise, some of these correlations are counterintuitive. For example, the results suggest 
that perceptions of increased community security are associated with normalization of certain 
acts. Specifically, those who perceive their communities as more secure are less likely to think 
that hitting someone and selling drugs are acts of violence. Although we cannot test this 
hypothesis further with the data available, one potential explanation for this result is that 
normalization of violence is intertwined with perception of security. In other words, respondents 
who believe that their community is safe (despite being objectively violent) are more likely to 
normalize certain acts of violence. That is, their normalization of violence could be biasing their 
own security perceptions. Relatedly, as we will review in the focus group analysis, we find that in 
some instances, youth classified selling drugs as a nonviolent act because gangs protect their 
communities, thus making drug activity a nonviolent act. But again, we cannot make causal claims 
about these correlations, and these hypotheses should be subject to future research. 
The survey results also suggest that victimization (which captures petty to severe crimes as 
described above) does not play a major role in the normalization of violence. Generally speaking, 
violence is more normalized among men. The effect of SES depends on the act in question. Violent 
acts (e.g., nonconsensual sex) are perceived as more violent as SES increases, and nonviolent acts 
(e.g., greeting or helping someone one doesn’t know) are perceived as less violent as SES 
increases. In the focus groups, there were interesting insights into this finding. Youth from lower￾SES backgrounds tended to find helping someone as dangerous because their intentions are 
unknown and there may be an intention to harm. We review this in more detail in that section. 
Violence is less normalized as age increases, and this is more salient across certain, more extreme 
acts, including nonconsensual sex, nonconsensual kissing, and hitting.
Table 2. Normalization of Violence, Insecurity, and Victimization
27
Table 3 examines the relationship between violence normalization and school victimization with 
the same controls as in Table 2. While victimization outside of school does not seem to play a 
major role in the normalization of violence, victimization in a school setting does play a role in 
certain circumstances. In particular, being touched leads to a decrease in the normalization of 
being hit on. Being robbed leads to a decrease in the normalization of hitting, although this effect 
is weak. 
Table 3. Normalization of Violence and School Victimization
Table 4 examines the relationship between violence normalization and experience with 
domestic violence, and Table 5 examines its relationship with interpersonal trust, again with 
the same controls. The data show that domestic violence (verbal or physical abuse at home) does 
not have a prominent effect on normalization of violence; however, there are some relationships 
of note. Verbal abuse in the home is associated with a normalization of nonconsensual kissing; 
physical abuse is associated with a normalization of criticism. We should note that, given the 
sensitive nature of these questions and the interview mode, underreporting of domestic violence 
is likely to be high and concealing of its potential impact on normalization of violence. Across the 
sample, verbal abuse is “somewhat” or “very” common across 20.1% of respondents, while 
physical abuse is “somewhat” or “very” common across 13.4% of respondents.
While trust in one’s neighbors is not associated with a normalization of violence, trust in one’s 
family is associated with a decreased normalization of violence across many actions (e.g., 
nonconsensual sex, hitting on, criticism, nonconsensual kissing, hitting, and selling drugs). As we 
discuss later, the focus groups provided more insights about the relationship among family 
relations, the experience of domestic abuse, and normalization of violence.
28
Table 4. Normalization of Violence and Experience with Domestic Violence
Table 5. Normalization of Violence and Interpersonal Trust
As part of our analysis, we also explore correlations between our violence strata with 
normalization variables and other important survey variables (see Section D in the Appendix). 
Overall, correlations seem to suggest that normalization of violence is more likely to be 
influenced by specific individual-level experiences and not so much by recent community-level 
violence. The correlations show that areas with high levels of violence do not necessarily 
correlate with higher victimization or normalization of violence. However, the analysis does show 
correlations between victimization and perceived security and, most importantly, between 
victimization across crimes and strong correlations between normalization of various elements
of violence. 
Protective Factors: Organizational Affiliation and Program Participation
In this section, we explore participation in different organizations and substance-abuse prevention 
programs, and whether this has any impact on violence normalization. We first explore some 
descriptive statistics regarding youth participation in relevant programs across states. We then 
show a series of regression analyses to explore the degree to which program participation may 
influence the normalization of violence among youth. Our results suggest that violence is most 
normalized among those who participate in certain organizations such as sports and community 
groups. These results are not causal conclusions (i.e., participating in sports does not necessarily 
lead to violence normalization), but suggest that those in such groups report higher levels of such 
normalization. We discuss these results in more detail below. 
29
Figure 5 presents the percent of youth participants who had participated in violence prevention 
programs, and Figure 6 presents the percent of youth participants who had participated in drug￾abuse reduction/prevention programs, both by state and gender. A total of 10.9% of males and 
9.7% females had participated in at least one violence prevention program, while a total of 17.3% 
of males and 10.4% of females had participated in at least one substance-abuse prevention 
program. We observe that in more than half of the states, male participation is more common in 
violence prevention programs, although there is a good number of states (16) where female 
participation is more widespread. In terms of substance abuse, we see that, in general, male 
participation is more common.
Figure 5. Percent of youth who have participated in violence prevention programs, by state 
and gender.
Given these relatively high participation rates in prevention programs, it is interesting to examine 
the relationship between organizational affiliation and violence normalization. Table 6 looks at 
the effect of current organizational affiliation and the normalization of different violent acts, with 
the same controls as before. The results show that the type of organization matters for its effect 
on normalization. We observe that affiliation with community organizations and, to a lesser 
extent, affiliation with sports are associated (although only at a 10% confidence level) with a 
decreased perception that nonconsensual sex is a violent act (this act is normalized), while 
affiliation with community organizations is associated with a normalization of hitting on an 
individual. Affiliation with a political party is associated with normalization of both criticism and 
selling drugs. These results are not causal conclusions (i.e., participating in a community 
organization does not necessarily lead to violence normalization), but suggest that those in such 
groups report higher levels of such normalization. This may suggest that interventions should be 
targeted at these youth. In the case of community organizations, for example, it may suggest these 
groups are appropriately reaching youth already at risk of normalization. 
30
Figure 6. Percent of youth who have participated in drug-abuse reduction/prevention 
programs, by state and gender.
The results also show that those who are affiliated with religious groups see selling drugs as more 
violent. Affiliation with student organizations decreases the normalization of nonconsensual sex 
and hitting (they are seen as more violent) at confidence levels of 10% and 5%, respectively. In 
general, it seems that participation in school organizations is associated with the most positive 
and consistent effect on normalization of violence. 
Table 6. Normalization of Violence and Current Organizational Affiliation
Gang Participation
In this section, we explore some descriptive statistics and show a series of regression analyses to 
examine how exposure to gangs, socioeconomic factors, organizational affiliation, and prevention 
program participation are associated with “gang risk factors,” or risk assessments regarding gang 
participation. Overall, we find that men typically know more gang members than their female 
31
counterparts, but this is also influenced by the security of the community (the more insecure one
perceives the community is, the higher the percentage of youth who know gang members). 
Furthermore, we find that those who know gang members consider joining gangs to be more fun. 
As SES and educational attainment increase, so does the perception that gang participation will 
result in death and injury. Participation in certain types of organization, such as sports, is 
associated with lower levels of these beliefs. 
Figure 7 shows the percentage of respondents who know someone in a gang, separated by 
gender and perceived security of one’s community. As can be observed, the percentages are 
much higher for males (32.4% of males vs. 22.4% of females), and the percentages increase as the 
perception of insecurity in one’s community increases.
Figure 7. Percentage of respondents who know someone in a gang, by gender and perceived 
security of one’s community
Table 7 examines the effect of knowing someone who belongs to a gang on the perception of 
risks associated with joining a gang, controlling for the same socioeconomic variables as above. 
Lower values indicate more agreement with a statement (e.g., joining a gang increases security), 
while higher values indicate more disagreement (e.g., joining a gang does not increase security). 
Results indicate that victims of crime tend to believe that gangs increase one’s security, but also 
increase one’s chance of dying; victims of crime are also less likely to believe that being in a gang 
will lead one to end up in prison. Those who know gang members personally are more likely to 
believe that gangs are fun, but this also increases the likelihood of physical injury. As SES increases, 
individuals are more likely to believe that joining a gang will result in death and injury. Males are 
more likely to see gang membership as fun. As age increases, individuals are less likely to believe 
that gangs are fun and more likely to believe that membership will result in death and injury. As 
educational attainment increases, individuals are more likely to believe that gang membership will 
result in death, injury, and imprisonment. 
32
Table 7. Gang Risk Factors
Expanding on the results above about protective factors and violence normalization, Table 8 
shows the relationship between affiliation with different types of organization and perception of 
risks associated with joining a gang. Respondents affiliated with community organizations are 
more likely to report that gangs are fun and secure, while underestimating the potential risks 
associated with gang activity (death, injury, and imprisonment). This underscores the potential 
role of social networks as enablers of gang involvement. In other words, youth involvement in 
gangs does not happen in isolation, and it is plausible that certain local community organizations 
serve as hotbeds for gang recruitment. Based on the available information, we cannot disentangle 
the specific types of community organization that may facilitate or be correlated with gang 
involvement. It is very likely that other types of organization or group activity disincentivize gang 
involvement. Note, for example, that participation in sports (which is weakly correlated with 
normalization of nonconsensual sex, likely because the proportion of males is much higher) is, in 
fact, associated with a decrease in the perception that gangs are fun and an increase in the 
perception that gang membership leads to death and imprisonment. Furthermore, it is also worth 
mentioning that current participation in student organizations has no relationship with perception 
of gang risks. 
We ran further analysis to explore the association between past participation in violence 
prevention and drug-abuse prevention programs on these perceptions (see Section E of the 
Appendix). We find that while past participation in violence prevention programs is associated 
with a decrease in the perception that gang membership is fun, participation in drug-abuse 
prevention programs is associated with an increase in the perception that gangs increase one’s 
security and are fun. Notably, this is not a causal relationship; it is possible that those who 
participate in drug-abuse prevention programs are more likely to hold these views in the first 
place.
33
Table 8. Gang Risk Factors and Current Program Participation
Overall, the results from the survey seem to suggest that normalization of violence is more likely 
to be influenced by specific individual-level characteristics and experiences and not so much by 
recent community-level violence. Individual-level characteristics such as gender, age, and 
socioeconomic background, as well as exposure to violence in one’s home, seem to matter more. 
The results also point to the potential role of social networks as enablers of normalization and 
gang involvement and that this varies by type of organization. This finding may suggest that youth 
involvement in gangs does not happen in isolation, and it is plausible that certain local community 
organizations serve as hotbeds for gang recruitment. At the same time, it is also very likely that 
other types of organization or group activity disincentivize normalization and gang involvement.
Based on the available information, we cannot disentangle the specific types of community 
organization that may facilitate or be correlated with normalization and gang involvement, but 
this is a relevant area of opportunity for future research. This is beyond the scope of our analysis, 
but more fine-grained data on individual exposure to different forms of community-violence over 
time would improve our understanding of the potential mechanisms that influence violence 
normalization. It is important to keep in mind that these correlations are not causal findings and 
should not be interpreted as such. 
V.2 Experimental Results
In this section, we explore the experimental results. We run two sets of models, both of which 
employed a linear model (ordinary least squares [OLS]). For the first set of models, the outcome 
variable is an index of blame capturing blame for the individual perpetrator in the vignette 
(Rodrigo). Respondents are asked how much blame they attribute to the perpetrator on a scale 
of 1 to 7. We treat this as a continuous variable. In these models, positive coefficients can be 
interpreted as more blame attributed to the individual, while negative coefficients indicate less 
blame attributed to him. 
For the second set of models, the outcome variable is an index of blame capturing blame for the 
individual perpetrator in the vignette (Rodrigo) vs. blame for society and the government. To 
calculate this index, we average the level of blame attributed to the society and government and 
subtracted this value from the level of blame attributed to the individual perpetrator. We then 
34
normalize this score to be from 0 to 1, where lower values (below 0.5) indicate more blame 
attributed to society and the government and higher values (above 0.5) indicate more blame 
attributed to the individual. 
In terms of the explanatory variables, we use the severity of crime, the upbringing of the 
perpetrator, his position in the gang, and the type of victim. The severity-of-crime variable is 
constructed as a continuous variable, with robbery, extortion, kidnapping, and assassination 
assigned values from 1 to 4, respectively. All other variables are categorical. Reference categories 
remain constant across all tables in this section: for class, the reference variable is “lower-class”; 
for type of victim, the reference category is “businessman”; for perpetrator position in the gang,
the reference category is “following orders.” All coefficients can be interpreted as comparative 
effects between the reference category and listed category.
In the first set of models, coefficients should be interpreted as the average increase in blame to 
the individual on the 1-to-7 scale compared to the reference category. For example, in Table 
11, if Rodrigo is middle-class (vs. lower-class), the estimated mean of blame attributed to him 
will increase by 0.215 points on the 1-to-7 scale. In the second set of models, positive coefficients 
indicate more blame attributed to the individual compared to the government and society, while 
negative coefficients indicate the opposite. In other words, the latter measure only tells us 
whether more blame is attributed to one actor over the other, but it doesn't capture the amount 
of blame being attributed to such actors. 
For all of these models, we run regressions using the pooled results (the whole sample), and we 
also divide populations by relevant categories, including the municipal homicide rate, municipal 
MRP victimization rate, municipal MRP perceived security, gender, and SES of respondents. To 
create the two samples for each comparative test, the total survey sample is divided in half 
based on the median value of the variable of interest (e.g., homicide rate).
Table 9 and 10 summarize the results found for individual blame attribution and for index of 
blame capturing blame for the individual perpetrator vs. blame for society and the government, 
respectively.
35
Table 9. Individual blame attribution
Variable/ 
Sample
Pooled High 
Homicide
Low 
Homicide
High 
Perceived 
Insecurity
Low 
Perceived 
Insecurity
High 
Victimi￾zation
Low 
Victimi￾zation
Male Female High SES Low SES
Middle￾Class
+ Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame
Crime 
Severity
(Increasing 
Value)
– Blame + Blame + Blame + Blame
Gang 
Leader
+ Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame + Blame
Victim 
(Worker)
+ Blame
Victim 
(Local 
Politician)
– Blame
36
Table 10. Individual vs. Government + Society Blame Attribution
Variable/ 
Sample
Pooled High 
Homicide
Low 
Homicide
High 
Perceived 
Insecurity
Low 
Perceived 
Insecurity
High 
Victimi￾zation
Low 
Victimi￾zation
Male Female High SES Low SES
Middle￾Class
+ Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo
Lower￾Class
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
Crime 
Severity 
(Increasing 
Value)
+ Rodrigo + Rodrigo + Gov & 
Soc
+ Rodrigo
Gang 
Leader
+ Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo + Rodrigo
Following 
Orders
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
+ Gov & 
Soc
Victim 
(Worker)
+ Rodrigo
Victim 
(Local 
Politician)
+ Gov & 
Soc
37
Table 11 presents the pooled results for the individual blame attribution model. What stands 
out are Rodrigo’s characteristics. According to our hypothesis, on average, Rodrigo is more to 
blame if he comes from the middle class compared to the lower class and if he is a gang leader 
compared to a member following orders. Contrary to our expectations, the severity of the crime 
is not significant, nor is the type of victim. This result is further explored in other models. 
Table 11. Blame for Individual, Pooled Results
When dividing the sample by relevant characteristics, we also find that, in general and consistent 
with the pooled results, Rodrigo is more to blame if he comes from the middle class. However, 
this result is not significant in low-homicide areas, meaning that in areas where respondents are 
rarely exposed to homicidal violence, Rodrigo’s upbringing does not impact the blame attributed 
to him. In other words, Rodrigo’s actions are evaluated asymmetrically depending on whether 
the respondent is located in a low- vs. high-violence area. One possible interpretation of this 
finding is that respondents in localities with lower homicide rates are more likely to be middle￾class and, therefore, evaluate Rodrigo’s actions in a more positive or neutral light. This 
phenomenon is often referred to as “intergroup bias,” which is characterized by a systematic 
tendency to evaluate one’s own membership group more favorably than the “out-group” 
(Hewstone et al., 2002). This could also mean that respondents in lower-homicide areas simply 
do not care about the perpetrator’s characteristics, and they assign blame equally. The results 
suggest that further research is needed to identify the mechanisms underlying such asymmetric 
behavior. It is also noteworthy that the baseline level of blame is higher in high-homicide areas 
compared to low-homicide areas. This can be seen by examining the constant in each model, 
which is about 5.1 for high-homicide areas and 4.7 in low-homicide areas. Further, we see that in 
high-homicide areas, crime severity has a negative relationship with blame, but there is a positive 
relationship with blame in low-homicide areas. However, the effect size for this variable is quite 
small in high-homicide areas, indicating that those from such communities may be less sensitive 
to differences in crime severity. In Table 12, we parse apart the sample by homicide level. 
Models parsing apart the sample by MRP security and victimization level are presented in Tables 
1 and 2 of Section F in the Appendix. 
38
Table 12. Blame for Individual by Community and Homicide Level
In the models that capture blame for the individual perpetrator vs. blame for society and the 
government, the pooled results (shown in Table 13) show that, in general, Rodrigo is more to 
blame vs. the government and society if he comes from the middle class, and, conversely, the 
government and society are more to blame if Rodrigo comes from the lower class. Further, we 
also see that if he is the leader of the gang, more blame is attributed to him, and if he is following 
orders, more blame is attributed to the government and society. 
Table 13. Blame for Individual vs. Government + Society, Pooled Results
However, when we parse out the sample by the level of violence in the respondents’ areas of 
residence, we see that the result of more blame attributed to Rodrigo (relative to the government 
and society) if he comes from the middle class only holds in areas with high homicide and high 
perceived insecurity (for the victimization breakdown, we found a null effect for this variable).
Table 14 presents the results for the sample broken down by homicide level; the models for 
insecurity perception, victimization, SES, and gender can be found in the Appendix. These results 
suggest that individuals in high-violence areas perceive perpetrators who are economically 
disadvantaged as less to blame for their actions; rather, society is at fault. It is possible that 
individuals perceive such crimes as “more” justifiable as the government has not adequately 
provided for lower-class individuals. Conversely, in low-violence areas (homicide and perceived 
39
insecurity) the results suggest that the perpetrator’s upbringing does not seem to matter when 
respondents assign blame to the individual vs. the government and society. These respondents 
do not evaluate differently someone who is relatively richer compared to someone who is 
relatively poorer. It is possible that respondents in these areas do not identify themselves with 
the perpetrators or that they belong to middle- or upper-class networks, in which case, their 
assessments of blame are less likely to be informed by the perpetrator’s upbringing. 
Table 14. Blame for Individual vs. Government + Society by Homicide Level
It is important to note that, regardless of community security level, respondents consistently 
perceived the perpetrator as more to blame if he is a gang leader; this suggests that the 
government and society are, generally speaking, seen has having little influence on the decisions 
of higher-ranking gang members, whose actions are based more on their individual decisions.
As mentioned above, the pooled results show that blame attribution toward Rodrigo is mainly 
about his personal characteristics, as the severity of crime has no effect. However, when we 
divide the sample by violence level in respondents’ areas of residence, we found that not only do
Rodrigo’s characteristics matter, but also what he does matters (i.e., the severity of the 
committed crime). Both the models on individual blame attribution and on individual blame 
attribution vs. the government and society show that Rodrigo’s blame increases with crime 
severity in areas of low homicide (see Tables 12 and 14), low victimization, and low perceived 
insecurity (see Tables 1, 2, 4 and 5 in Section E of the Appendix). These results suggest that, 
in areas that are not as exposed to crime, individuals are more to blame for more severe crimes. 
It is also notable that the level of blame attributed to Rodrigo decreases with crime severity in 
high-homicide areas, and, while the coefficient estimates are not significant for other sample 
breakouts (although they have the same sign), the government and society are more to blame for 
more severe crimes in high-victimization areas. As before, this result may suggest that individuals 
who are not exposed to crime do not identify with the perpetrators (as opposed to areas with 
high violence) and, therefore, blame the individual more for more severe crimes. This result may 
also suggest a perception that the government is not protecting against severe crimes in high￾violence areas. Finally, the results seem to suggest that the blame attribution depends on the 
context in which one operates. In higher-homicide areas, people see crime involvement as part 
of organized crime; therefore, homicide is part of the nature of violence. More severe crimes are 
not unexpected; therefore, respondents placed less blame on the individual for these types of 
40
crimes, which are more common. It might be that this is the way disputes are usually resolved, 
which speaks to crime normalization.
Overall, there seems to be a class divide shown by the results broken down by both respondent 
context and Rodrigo’s identity. The average attribution of blame for Rodrigo is higher in areas 
with higher levels of insecurity/homicide/victimization. However, respondents’ assessments of 
blame are more sensitive to the type of crime in low-violence areas. Respondents who live in 
safer areas are more likely to increase the amount of blame to Rodrigo if the crime is more
severe. This could be caused by the fact that severe crimes, such as homicide, stand out as 
extraordinary or spectacular forms of violent criminal behavior—and are therefore less likely to 
be condoned than less severe crimes—whereas in high-homicide areas, either severe crimes have 
been normalized, or criminals have been heavily stigmatized and perpetrators are thought to 
engage in all sorts of criminal behavior (if Rodrigo engages in extortion, it is very likely that he 
has committed more severe crimes). 
V.3 Focus Groups
As mentioned in the methodology section, focus group participants were young offenders (with 
the exception of those from PILARES) from poor socioeconomic backgrounds who had
participated in reintegration programs. This means that our participants had been involved in 
violence and also had gone through processes to help them understand what happened, recognize 
automatic behaviors, and revise their choices while working to devise better paths for the future. 
Throughout the focus group discussions, we find that youth oftentimes have mixed or even 
contradictory views about violence in all its forms, and therefore, all of the results presented 
herein may portray a tension between the different arguments expressed by youth. We consider 
this to be a reflection of the reality in which youth struggle to reconcile past—mostly harmful—
choices with opportunities for a better future. These are incredibly difficult mental processes in 
which youth have to understand what drove them to make those choices, understand their 
contexts, and work very hard to move forward. The results we present in this section reflect this 
reality, accompanied by tensions in different rationales with which youth struggle.
To further understand how complex these processes are, it is important to remember the 
context from which these youth come and how their beliefs, views, and values are shaped by 
their environment. Here, we understand beliefs as what is considered true based on assumptions 
and past experiences. Views point to someone's attitudes when assessing a situation, and values 
indicate what is important when making decisions. This environment also determines automatic 
behavior, which can drive negative outcomes such as violence or delinquency involvement, 
because people often respond to a context without conscious deliberation. While automatic 
responses are generally adaptive, they can also be ill-suited. Disadvantaged youth face greater 
situational variability, therefore increasing the likelihood that automaticity leads to negative 
outcomes (Heller et al. 2017).2 We consider that this situational variability impacts all sorts of 
youth reflections about what is right and wrong, what is acceptable, and what is understandable.
2 Heller et al. (2017) explained this situation with a very clear example: Consider two situations faced by youth in 
which they have to deal with different forms of authority. At school, teachers assert authority by telling students to 
be quiet. On the street, a larger person may demand money or their mobile. The adaptive response of a middle￾class youth is to comply. They will do what the teacher says and hand over their phone and then go and tell authorities 
41
In the analysis below, we first describe the sociodemographic characteristics of respondents,
further expand on the survey results, and explore how precarious work conditions influence 
youth motivations and aspirations. Second, we explore how youth normalize violence and what 
mediates this normalization. We also analyze in detail certain mediators of violence normalization 
that were found as highly relevant through the focus group analysis, namely gender-based 
violence, domestic violence, substance abuse, and social norms. Third, we investigate the 
rationales behind the advantages and disadvantages to joining a gang as perceived by youth. 
Fourth, we explore how protective factors as enabled by prevention programs may prevent 
recidivism, and we identify important areas of opportunity. Finally, we analyze the results of the 
vignette as presented in the focus groups, allowing us to further disentangle mechanisms behind 
blame attribution as found in the survey. 
Sociodemographic Characteristics of Respondents
The focus groups confirmed some of the survey findings and enabled further exploration of how 
violence normalization and blame attribution take place among at-risk youth. As mentioned in 
the methodology section, focus group participants were a segment of youth who had been 
exposed to various types of victimization and who are from low-SES backgrounds (see Section G 
of the Appendix for a distribution of reports of domestic abuse by SES). Their neighborhoods are 
insecure; fights between neighbors and drug dealing are part of the daily scenarios in which these 
youth live. The presence of groups with violent tendencies is also frequent, such as gangs and 
junkies living on the street. This is particularly marked in Ciudad Juárez, as explained by one 
participant from La Tenda:
“ … in my neighborhood, in the street I live in, there are some that fight all the time. It’s a family 
of junkies … When I was younger, once there was a ‘tecato,’
3 skinny, malnourished, without 
having eaten in days, and jarocho. A muscled guy came, younger than him, grabbed him, shook 
him by his shirt, lifted him, and ‘poom,’ down to the floor. He hit his head and started bleeding, 
horrible … ”
As mentioned above, all groups convened young offenders, with the exception of the PILARES
group. Participants from PILARES were at-risk youth who live in low-SES neighborhoods in 
Mexico City, but who had not been involved with criminal activity and who have had more 
protective factors than youth from the other groups. These participants had, on average, a higher 
educational attainment compared to the rest, as most had finalized their upper-secondary 
about what happened. For a youth from a more disadvantaged background, school requires compliance, but on the 
street, it may be adaptive to develop a reputation as someone who will fight back to deter future victimization. 
Handing over their phone may create a signal of weakness. Automaticity interacts with the social environment. 
Disadvantaged youth face heterogeneous situations that demand different responses. While middle-class youths can 
simply comply in both situations, for disadvantaged youth, if they comply in both situations, they will be abused on 
the streets. If they resist in both situations, they will do poorly in school. The situation requires different responses 
that create automaticity more costly for the disadvantaged youth. Disadvantaged youth have to learn to recognize 
when not to be automatic. 3 Tecato: slang from Ciudad Juárez meaning a drug addict who mostly consumes crystal meth and crack by inhaling 
from a Tecate (beer) can.
42
education and some are in college, while for all the other CSOs, participants were either finishing 
middle school or studying at the upper-secondary school, if at all. 
The focus groups gave further detail into the type of jobs to which youth have access and the 
consequences this has on their motivations and aspirations. In general, the jobs they get are 
precarious. Jobs are usually sporadic, without social security, a contract, or a stable salary, and, 
in most cases, youth from lower socioeconomic backgrounds are self-employed (in very low-pay 
activities). In terms of income, participants expressed it not being sufficient (the highest salaries 
were found in the manufacturing industry at around 500 USD per month). Some PILARES
participants (who, as mentioned before, have slightly higher SESs), reported having access to 
better jobs. For instance, one outlier was a youth working for a company as a community 
manager. 
The focus groups also confirmed the differences in youth employment by gender. In most groups, 
youth mentioned that most women in their neighborhoods do not work or study. When they 
do, it tends to be in highly stereotyped occupations. One notable finding is that for this population, 
early pregnancy was mentioned as a common reason for school abandonment, as well as the 
desire of young women to be in a partnership at an early age. Table 15 synthesizes youth 
occupations. 
Table 15. Youth occupations
Program description Women Men Both
Employed
Cook
Security staff in a bar
Mason
Butcher
Nurse
Tailor
Catering
Maquila
Factories
Call center
Self-employed
Box trainer
Deliverer (Rappi, 
restaurants)
Loader at market
Stand builders at air 
markets
Uber
Mechanic
Temporary migration
House-cleaning
Nails
Selling catalogs
Selling clothes (by internet, 
air market, house)
Blacksmith
The focus groups gave us a better sense of youth aspirations and motivations derived from this 
precarious job context. It was common for participants to express that having a job they like is
unattainable, and therefore, being employed with an income is enough. Given these perceptions, 
a lack of a personal projects is common among participants. The most common aspirations are
establishment of different types of small business (i.e., food stand for males and tailoring and 
catering for females) and self-employment as tattoo artists, professional football players, and 
musicians for males. Commonly expressed desires are to be their own boss, to have control of 
their own time, and to have enough to make a living, in contrast to jobs like “maquila” in the 
north and call centers in Mexico City. Only a few expressed looking forward to continuing 
studying in college and becoming professionals. Moreover, some youth from the north see 
migrating to the US as the only option, where they think they can access better job opportunities. 
43
Others mentioned that it is very common for youth to have the aspiration of engaging in criminal 
activity, as this is the most promising path to become famous and to gain respect.
Violence Normalization
Types of Violence Normalization
Through the focus group discussions, we identified four scenarios of violence normalization. The 
first is when violence is not normalized at all; that is, participants recognize it and actively reject 
it. Then, we categorized three types of violence normalization based on the extent of 
normalization, defined as nonrecognition, habit, and acceptance. The nonrecognition of violence is 
the strongest form of normalization and happens when an offense or aggression is not perceived 
as violence, but as some other form of social interaction. The habit of violence is when violence 
is recognized but is seen as a habit, as something that “everybody does.” The acceptance of 
violence is when violence is perceived as inconvenient but necessary to solve problems. These 
forms of normalization are mediated by the automated responses of youth, as well as their beliefs, 
views, and values related to gender, emotional regulation, substance abuse, social norms, 
domestic violence, type of aggression, and perception of threat or harm from others. 
While participants reject violence most of the time, they do normalize violence in specific 
situations. While to some youth certain forms of violence might be unacceptable, other forms of 
violence can be necessary, and some others may not even be considered violence. We found 
some tensions between the different rationalities presented by youth while discussing violence, 
which, as mentioned above, is a reflection of the realities in which they live and the contradictions 
they have to manage between what they are or were exposed to and how to build better 
opportunities for themselves. 
Table 16. Violence normalization categories by mediator
Mediator / 
category of 
violence
No violence 
recognition
Habit of violence Necessity of 
violence
Rejection of 
violence
Gender 
stereotypes and 
misogynist views
Partner relationships
Women’s 
competition for men
Flirtation/jealousy
Men are protective
To gain respect Sexual 
advances 
without 
consent
Intimate 
partner 
relationships
Economic 
control
Substance abuse
Rape while drunk
Selling drugs
Fights while high
Fights while drunk
Persistent 
aggression
Violence 
increases with 
substances
Rape while 
drunk
Harm to 
consumers
44
Domestic 
violence
Verbal abuse Partner relations Parenting styles
Partner relations
Economic 
control
Social norms Aggressions 
among friends
Bullying Peer pressure
To gain respect
Perception of 
threat or harm
Verbal and 
psychological 
abuse
Excess of contact 
(sports, transport)
Bullying
Persistent 
aggression
Debts
Work
Self-defense
Defense of others
Lack of respect
Death
Pregnancy
Psychological 
harm
Emotional 
regulation
Impulse To calm others It might 
escalate
No recognition of violence
No recognition is the strongest type of violence normalization and relates to some participants’ 
perceptions that actions not entailing physical aggression, such as verbal or psychological violence, 
are not violent as they consider there is no real harm. Similarly, some of the participants’ 
responses indicated that verbal abuse and certain forms of psychological aggression (e.g.,
criticizing and insulting) are not seen as violent when they take place among friends. Although 
infrequent, one noticeable result from the women’s section of La Tenda di Cristo in Ciudad 
Juárez is the belief that rape is not violent if there is no resistance from the victim, such as when 
she is drunk:
“I don’t see it as something violent, I see it as something that shouldn’t be done and shouldn’t be 
forced, but it is not violent … when you are drunk, you don’t struggle.”
Another noticeable result is the perception that selling drugs is not just not violent, but that it 
can be beneficial, as criminal organizations protect those who live in the areas where they sell 
(replacing the police). As mentioned in the survey analysis, an unexpected but interesting finding 
is that as community security perception increases, so does the normalization of selling drugs, 
which might be partially explained by this mechanism. This is illustrated by one participant from 
Reinserta in Mexico City:
“ … where I live, drug-dealers are the ones who protect us. I live in a housing unit. Outside, there 
is a huge fence, there are many entrances, but only those who live there get in. When someone 
gets inside to steal cars or something else, they take action. For example, my stepfather works 
with Uber. He leaves his car outside, and when they have tried to steal something from it, those 
who act are the drug-dealers, and they do it in their way. Before, they called the police, but it 
took them centuries to arrive. So, it’s the drug-dealers who protect us. There might be some 
violence, but I don’t think it’s too much.”
45
Habit of violence
The second type of normalization—when violence is recognized but conceived as a habit or 
something that “everyone does”—is frequently related to gender-based violence, such as in 
intimate partner relationships, and to gender stereotypes that dictate that men have to be 
protective and dominant, that women should compete for men, and that men can be violent when 
jealous or flirting:
Participant 1: “I have girlfriends where I live who are abused by their husbands. Abuses, 
harassment, that’s normal. Yes, that’s very normal here.”
Participant 2: “As [Participant 1] said about harassment, violence starts with relationships, there’s 
no way that a stranger hits a woman just like that.”
This type of normalization also appeared when referring to substance abuse: it is expected that 
drunk or drugged people become aggressive and get involved in fights. Other cases relate to the 
type of aggression, such as in soccer games or in a crowd on the subway, for which participants 
pointed out that an excess of physical contact will understandably drive to violence. In a similar 
vein, bullying seems to be something that everybody does at school; although it is not considered 
good, it is just expected to occur. 
Finally, this type of normalization seems to be associated with emotional self-regulation and 
automatic responses. Some participants mentioned that violence can stem from frustration and 
anger, and although that is something negative, it just happens:
“ … the truth is that you get angry when people trick you, and sometimes that impedes you from 
controlling yourself and you act violently or by impulses. You should dominate them. My opinion 
is that it isn't good, but sometimes the impulse is stronger and you act with violence.”
Necessity of violence
Necessity of violence, which is the weakest form of normalization and happens when violence is 
considered negative but necessary in unbearable situations, is mostly related to perceptions of 
threat or potential harm against oneself or others, mainly one’s family. Also, responses that fall 
in this category referred to situations in which participants are aggressed first. Here, violence is 
considered necessary for self-defense or protecting others. This finding is consistent with the 
literature of automaticity. In more disadvantaged areas, it is considered adaptive to avoid conflicts 
(because they can escalate quickly), but predicts higher retaliation (Heller et al. 2017):
“I had a fight with Teeth, a guy at school. I don’t know what he wanted. Others were telling him 
to bother me and I was telling him, ‘Control yourself!’ Then, I told him, ‘We will fix this outside,’
and we fought. He was a foolish kid. I was telling him, ‘You think he’s going to defend you, but 
he only wants to see a fight,’ but he didn’t control himself. Outside of school I made him 
understand [by hitting him], just to see if he would keep listening to the others.”
46
“[You can use violence] just to solve problems because it’s been days, weeks. You can have a 
problem for several weeks and you want it to end. I know people who do that [violence] to 
terminate it, and that’s how the problem ends.”
Relatedly, some participants mentioned that the use violence is sometimes necessary to collect 
debts and in certain situations at work, such as when dealing with difficult clients. Finally, an 
interesting finding related to this category is that violence can be assumed to be necessary as part 
of the process of raising children:
“It is better to talk, but when they don’t obey, I’m not saying it is good to spank them, but when 
they don’t obey, I think you should. It is not hitting them as if you were going to kill them. As I 
say, I talk to my kids all day long, ‘Sit-down,’ ‘Be quiet,’ ‘Stop jumping,’ ‘Don’t go out.’ That’s how 
I spend my day. But when I see that it isn’t working, I just spank them.”
Rejection of violence
Although all groups manifested some type of violence normalization—with the exception of the 
PILARES group, which had an overall lower level of normalization—there were several 
statements against violence. In most cases, these statements stemmed from perceived harm 
associated with certain conduct. Here, both verbal and physical aggressions are mainly considered 
violent because they result in psychological or physical harm, including death. Also, certain 
comments pointed out that selling drugs is violent because it harms consumers and drug￾trafficking entails violence. 
Some responses also indicate a sense of moral harm related to verbal abuse and sexual 
aggressions by highlighting that they constitute a “lack of respect.” This includes one noticeable 
result from the women’s section in the Supera group: pregnancy being the main reason that rape 
is worse than any physical or emotional harm. Many participants stressed the importance of 
consent in any sexual advance, and consider that nonconsensual sexual intercourse while drunk 
is rape. Similarly, many participants reject intimate partner aggressions and highlighted how 
gender roles result in different forms of violence against women, such as economic control and 
household duties (this was more salient in the women’s section of the PILARES group). 
An additional and interesting finding is that many participants perceive that violent or criminal 
acts should not be perpetrated mainly because they are illegal, not because they harm others. 
This result might be related to their own risk assessments, given that these youth had already 
been in touch with the justice system, with the exception of the PILARES participants. Relatedly, 
we found that a common mechanism toward violence rejection is risk assessment. Youth reject 
violence because it can escalate. Again, this might relate to their own previous experiences, given 
that these youth had been involved with violence in various spheres (more than once) and were 
also participants of violence prevention programs that raise awareness about these issues. This 
type of assessment was frequent, but not exclusive, during discussions on substance abuse, where 
many participants mentioned that violence tends to escalate when people are drunk or drugged.
47
Analysis of selected mediators
Gender stereotypes and misogynist views 
While in general all groups recognized that women suffer more sexual violence and intimate 
partner violence than men, the focus groups also disentangled a layer of normalization of violence 
that is gender-based and rooted in gender stereotypes and misogynist views about how women 
and men should behave and their respective roles in society. Also, the discussions reveal that 
gender-based violence is prevalent in all social spheres where these youth participate—
community, work, school, friends, family, and intimate partner relationships—which conflicts with 
the recognition of violence as such.
Participants identify different forms of intimate partner violence, including physical, verbal, and 
emotional abuse, economic violence, and abandonment. As explained by a male and a female 
participant,
“I have seen cases where men don’t accept that women earn more than them. They say, ‘Fuck, 
this bitch is bringing more money than me, no way.’ That’s one reason why violence against 
women begins. As we were saying earlier, because of the crisis, women have to work, and men 
start imagining things such as, ‘I’m not earning that much. She might get to know someone else 
who earns more, I shouldn’t let her go to work.’”
“When my father abandoned us, my mother had to show what she was made of. She struggled, 
she struggled a lot.”
Youth also recognize how this violence can be transformed into physical abuse (inside or outside 
the home) that oftentimes ends up in homicide. It is interesting to note that the recognition of 
violence against women was much more significant among males in Ciudad Juárez relative to 
other places. One potential explanation for this finding is the context of Ciudad Juárez, which is 
one of the places with more violence against women in Mexico. During the past decade, a 
movement to increase visibility and address this violence has grown, and this topic is covered in 
some of the prevention programs in which the youth had participated. Moreover, these youth 
live near where the “Muertas de Juárez” are buried, which makes gender-based violence visible 
to them. As two participants mentioned,
“I say women (suffer) more, no? Yes, it is women, there have even been more deaths and 
assassinations. It might be the crisis, because of coronavirus, there’s no work and men release 
their anger against their families. Young women are kidnapped.” 
“[Women suffer] abuse, they get stared at, or they can even be taken away—that is what my 
girlfriend says. One time a man in a car was whistling at her, telling her to jump into the car, and 
she just ran away. That happens very often, when the kidnapping of women and all of that. 
Sometimes they also get touched … I have seen that various times … Not only once, there was 
this girl in between many crazy men, for things like this they are affected. Even their own family 
touches them. It is something terrible.”
As mentioned above, the discussions disentangled a layer of normalization of violence that is 
gender-based and founded in misogynist views and gender stereotypes. In many instances, 
48
participants began by stating that both women and men “suffer equally.” As the discussion 
progressed, they tended to distinguish differences in the forms of violence that affect them, 
although considering them equivalent. The rationale behind this is that both women and men 
perpetrate violence against each other in their own ways, resulting in a sort of “victimization 
balance.” In the words of a male participant,
“There’s a certain balance in that. Women are more victims of physical violence than men 
because women can’t bear too much physical violence, but men are more victims of verbal 
violence than women. That’s something we all have known. By genetics, men are stronger than 
women, although there are women who are stronger than me, that’s true. However, normally, if 
you take two people from the same age, a man is stronger than a woman, but mentally, women 
are stronger than men. Men are more psychologically harmed and women are more physically 
harmed. Hence, I don’t feel that one suffers more violence than the other. Both suffer violence. 
Although both can suffer from both types of violence because there are women who rape men 
and people laugh at it, and there are women who feel bad just because a man told them they 
are ugly. Both are victims of violence, it’s not that one gets more.”
Interestingly, this rationale was partially shared by women in all groups with the exception of 
PILARES. Although they pointed out that women tend to suffer more sexual aggression and 
physical violence than men, they highlighted that there are men who are also victims of violence 
from their intimate female partners and that women are perpetrators of violence against women 
as well, which ends up balancing the victim-perpetrator ratio. This was illustrated by the 
discussion in the women’s section of the Consejo Ciudadano group:
Interviewer: “Thinking about your female friends and other women you know, which forms of 
violence do women suffer more?”
Participant 1: “Harassment, critics, there are women who do plots, women can be real bitches. 
Men also do things, when they tell you, ‘Come here,’ and they force you, they rape you … ”
Participant 2: “When your partner wants to dominate you.”
Interviewer: “Which other forms of violence do women suffer?”
Participant 1: “There isn’t that much anymore.”
Participant 2: “Did you read on Facebook that they tried to kidnap a young woman? … A man 
was driving the car, a woman was the copilot, and there was another woman in the back. The 
two women tried to kidnap the young woman.”
Participant 1: “They hit and scratch her … I think we are the weaker sex indeed.”
Interviewer: “Which forms of violence do men suffer more?”
Participant 1: “There are many who are hit by their female partners.”
Participant 2: “Yes, I agree, maybe not as frequent, you see that a little more with women, but I 
think it’s the same.”
Participant 1: “I think the same. Those women hit men, they do the same as a man does, they 
don’t let them go out either.”
Another relevant finding gathered from the men’s discussions is the perception that men are 
often victims of women’s manipulations and provocations, as they perceived women to have all 
the governmental and legal support, mainly regarding sexual offenses. For instance, to some 
participants, women push men to have sex, and then they manipulate and say it was rape because 
the law favors them. Other misogynist views point to women taking advantage of their gender to 
49
obtain things, such as getting a job, and to women being responsible for the offenses they 
experience on the street because of the way they dress. Many male participants also consider 
that men are victims of women’s movements (e.g., referring to supporters of the feminist 
movement as “feminazis”). This is accompanied by the perception that “feminism” is the women’s 
equivalent of “machismo” and that feminist movements are the reason why men are perpetrating 
more violence against them, as explained by these male participants:
“I think violence against women is getting harder because of the feminazis’ movement, feminists, 
who are revealing all that men have done to women over time. Obviously, as a man, I believe it 
is wrong that most men do that to women, but there has to be an agreement to reach equality, 
so nobody is over the other.”
“ … violence against women began with ‘machismo’ and violence against men is starting with 
feminism. Machismo and feminism are the same type of movement, they don’t seek gender 
equality, they want to dominate the other.”
Domestic violence 
Although the survey results found that domestic violence (verbal or physical) does not have a 
prominent effect on the normalization of violence, they did depict some correlations of interest 
such as the effect of verbal abuse at home with a normalization of nonconsensual sex. In that 
analysis, we hypothesize that given the sensitive nature of these questions and the interview 
mode, underreporting of domestic violence is likely to be high, therefore concealing its potential 
impact on normalization of violence. Moreover, the survey found that trust in one’s family is 
associated with a decreased normalization of violence across many actions (e.g., nonconsensual 
sex, hitting on, criticism, nonconsensual kissing, hitting, selling drugs), which may suggest that 
family dynamics are relevant in mediating violence normalization. 
In this view, one notable finding from the focus groups is that participants are highly exposed to 
violence in their families, and they point out that these violent dynamics are transmitted from 
one generation to the next. Youth claim that violence is learned through socialization from family, 
and that if someone grows up in an environment where violence prevails, it is likely that he or 
she will replicate it. In the words of some participants,
“I see what happens at home, it's highly probable that I follow the same path. For example, if 
my father hits my mother … When we are kids, we learn what parents teach us. At school, they 
teach you things, but you bring the values of your family, and if you see things that are not okay, 
it's highly probable that you will replicate them.”
“ … it depends on who you grow up with. For example, in my family, my father didn´t give a 
fuck. Every time the police passed by, he told my brother to yell insults at them. Then, when 
women passed, he told us to yell ‘mamacita’ … ”
Discussions also indicated that, as expected, gender-based violence is intertwined with domestic 
violence, as described situations reflected misogynist role models who reproduce gender 
stereotypes and drive violence. Some of the participants’ experiences suggest that gender-based 
violence is learned at home, which plays an important role on how normalization takes place. 
This was illustrated with two examples. One participant minimized violence when describing an 
50
episode of abuse by his father, and another one pointed out that most of the people he knows 
reproduce these dynamics, indicating that it is something that everybody does: 
“In my family, there is no violence. My father drinks and all, sometimes he gets a bit annoying, 
but he goes to bed. When I was younger, he would yell and smash dishes. I never saw him hitting, 
but once I saw him giving a slap to my mom, a small one. That’s when I grabbed him and kicked 
him out. That’s how violence was, just yells and smashed dishes.”
“ … almost everyone who lives up here: ‘My father hits my mom,’ and almost everyone I know. 
Well, I identify myself, my father used to hit my mom. It’s typical from here, I would say almost 
60% of the people I know.” 
Substance abuse
Substance abuse is frequent among participants and their social circles, including their families; it 
is part of their social dynamics. Access to drugs is easy for them; they can buy them in their 
neighborhoods, as they know where the selling points are in bars or at parties. Participants from 
PILARES mentioned that drug distribution occurs at schools as well. The most used drugs are 
cocaine, crack, marijuana, and, in the north, crystal meth. In all groups, alcohol abuse is frequent,
and going to parties to get drunk is a very common weekend activity:
“[We go to parties] … you just go to drink. My friends just go out to parties to drink. They don’t 
even dance; they just drink.”
Groups from Ciudad Juárez and Supera revealed a more intense use of drugs; some of them 
openly mentioned being active consumers, although they are participating in harm reduction 
programs with their CSOs. This is relevant given the dependency generated by different types of 
drugs and their different relationships with violence and crime involvement; for instance, there is 
wide evidence on methamphetamine increasing violent behavior (Brecht and Herbeck, 2013). 
This was further explained by two participants from CASA: 
Participant 1: “ … cocaine and crack alter you a lot. Some people use that type of drug to do 
things that are not good, or just because they want to have more and they don’t get the money, 
they hit you.”
Participant 2: “Those are the most vicious, they sell things from their own homes or pawn them.”
Participant 1: “Or stealing for drugs.”
Participant 2: “They end up killed for not paying.”
As mentioned before, substance abuse is a mediator of violence normalization. As expected, 
participants expressed perceiving that drugged individuals engage with violence or escalate 
violence. However, discussions also showed that substance abuse among family members relates 
to domestic violence. As with misogynist views and gender stereotypes, substance abuse is 
entangled with many of the family dynamics of these youth and plays a major role in generating 
violence among their relatives. This was illustrated by participants from Consejo, La Tenda, and 
Reinserta:
51
“At home, there are problems with my brother because of drugs, violence, but only that. For me 
it’s very stressful. The thing is that my brother stresses me out when he’s drugged. There are 
fights, and my father and my mother fight for the same reason.”
“For instance, it happens when a mother is an addict and has children, as happened to one friend. 
She has children and starts hitting them because of the ‘malilla.’4 When she stops drugging 
herself, the ‘malilla’ is a bitch, it’s tempting. I do hit my brothers, whoever is in front of me.”
“I know a girl whose mom is an alcoholic, and every time she would come drunk, she would kick 
her out of the house, she would tell her things and that kind of thing.”
“Women suffer from physical and psychological abuse. I once heard about a guy that was in his 
home and his father arrived very drunk. The mother tried to calm him down, but the father was 
very mad and it escalated to hitting. When the son arrived, he saw and the father hit him too.”
Social norms
Social norms mediate violence normalization by shaping the behavior of these youth through 
expectations, rewards, and pressure from their peers, communities, and society. From a broader 
perspective, youth recognized that society as a whole reflects “thoughts and beliefs” that 
“normalize everything,” in the words of one participant from Renace in Monterrey. Social norms 
were pointed out more than once as reinforcing conduct that drives violence, such as what is 
necessary to gain respect in their neighborhoods, as explained by one participant from Reinserta 
in Mexico City:
“You don't always run with the same luck. Society tags you and that's how everything starts. It's 
rare to see a guy selling candy and going to the countryside. Why? Because he's scared that the 
others will laugh at him or that they will put him down. Before, I was ashamed because, after all, 
they can laugh at you if you take the bus to sell candy. I thought those guys were going to humiliate 
me, they are going to say now you're selling.”
Furthermore, the need or desire to belong to a peer group by adapting to gain acceptance or to 
avoid being victimized was identified as another driver of normalization. This factor appears to 
be mostly related to bullying and other forms of more severe violence that also take place at 
school (although not exclusively), such as obliging someone to consume drugs or threatening
others with exclusion. These forms of peer pressure were explained by participants from 
PILARES and CASA:
“I had a group of friends that smoke, well, consumed drugs. They wanted to include a friend and 
told him, ‘If you don't consume, we will take you out of here.’ That group was considered popular, 
the school populars, and they wanted to kick him out if he didn't consume. He had to consume.”
“In the end, if you don't want to leave that social circle because you are insecure to stay alone. I 
mean, you may have chances to have friends, but for not going out of your comfort zone with 
your friends, you decide to start consuming and staying with them instead of choosing another 
group.”
4 Malilla: slang word for craving.
52
“When you arrive at school, you look for the group of those who break the rules, the ones that 
have control, you adapt to them. Thus, you look at a guy who is there, sitting alone, quiet, and 
you are going to start calling him names, teasing him, and that's how it starts, that is bullying in 
schools.”
The survey also found that nonviolent acts (e.g., greeting or helping someone you don’t know) 
are perceived as more violent as SES decreases. The focus groups allowed us to unpack this 
finding. Youth expressed that helping a stranger is dangerous because their intentions are 
unknown and there may be an intention to harm. This perception stems from their experiences
with their social environments, where a general mistrust in others prevails. These youth live in 
insecure areas, where crime and violence are expected. The rationale behind these answers was 
exposed in this comment: 
“I can greet someone I don’t know, but you don’t know how that person will react … where I 
live, I know nobody and I have seen that if an outsider comes and greets them, they will ask,
‘Who are you? Where are you from? Are you from here?’ They will take offense because they live 
with the expectation that something is wrong. They don’t know if the person they’re greeting is 
down the wrong path, if it’s the police or someone else. Obviously, they mistrust everyone.”
Gang Participation
The focus group analysis allowed us to further analyze the rationales behind the disadvantages 
and advantages of joining gangs as perceived by youth. 
Participants’ reflections on joining a criminal organization revolved around three factors: 
economic benefits, coercion, and protection. These are at the core of their risk assessments of 
the convenience of joining. Groups revealed different rationales in this regard, some against and 
others in favor. 
Among those against joining a gang, one rationale is the “entrepreneurial” one, which is that it is 
better not to join an organization if you are already making good money by yourself, as you would 
have to share your earnings. This rationale is exclusive to Mexico City’s groups and reflects the 
dynamics of criminal organizations. Compared to Northern Mexico, drug cartels’ territorial 
control in Mexico City does not seem to entail an absolute monopoly of criminal activity in the 
places where these youth live and travel; thus, there is more room to engage in criminal activity 
alone. In the north, youth considered as part of their risk assessment the potential reprisal they 
may receive if involved in petty crime in territories that are controlled by the cartels. 
Others pondered more risks than advantages, mainly related to getting killed and “losing 
freedom,” because once one joins, one cannot leave, which can also put their families at risk. 
These are the reasons why they described joining as inconvenient or undesirable. However, one 
interesting finding from Chihuahua’s groups is that participants did point out these risks, but did
not perceive them as enough reason to abstain from joining because, for them, earnings would 
compensate for the risks, and they would be able to give more to their families, even if they were
killed. 
53
This relates to another rationale, which is the “aspirational” one. This is the perception that 
joining an organization will enable both access to more profitable criminal activity, such as 
hijacking retail stores, and the possibility to make a criminal career by progressing in the 
hierarchy, as explained by one participant: 
“Profits increase … you aren’t going to enter a small house to steal a TV, a stereo, or a washing 
machine anymore. When you join, they are going to send you to places with more money and 
merchandise to steal, such as ‘Del Río,’ ‘Oxxo,’ or something like that. So it’s more what you’re 
going to get, and more what you share, since it’s not fifty-fifty, but it’s going to be more than what 
you had before.”
This thinking is reinforced by a “fear of coercion,” which was mostly found in groups from 
Northern Mexico, where drug cartels have greater control over the social fabric, determine who 
can perpetrate a crime in a given area, and take retaliation against those who are not aligned with 
them. As explained by one participant in Ciudad Juárez:
“There are two options: whether you join or not. If you don’t, there’s a risk that they will retaliate, 
that they say, ‘This dude didn’t want to join us, I should kill him, so there's less of an asshole to 
bother me.’ As they say, if you join, it’s going to benefit you and your boss.”
This view is also strengthened by the perceived “protection” from both the police and other 
criminal organizations. The first form of protection is dependent on the type of arrangements 
held by the organization with the police, while the other one relates to the amount of control 
had by an organization over a certain territory. As illustrated by one participant also from Ciudad 
Juárez:
“Sometimes it’s convenient because, like me, as I live up there in the last houses, there was a 
time when I was selling, and some guys came saying, ‘We’re going to take your brother to the 
sierra even if he doesn’t want to.’ ‘No, you aren’t going to take him. I know this guy.’ ‘ Call him 
to see if it’s true’. That’s how it helped, because my brother did want to go. It suited me to be 
with them at that time.”
Blame Attribution
Findings from the vignette discussions in the focus groups confirmed the results from the survey 
while allowing us to further disentangle mechanisms. As mentioned in the methodology section, 
the vignette was presented as an evolving story. It started with depicting Rodrigo as a youth from 
a low socioeconomic background who got involved in petty crime (by invitation from his 
neighbor) to help his mother. At this point, we inquired about the type of victim having any effect 
on blame attribution. Then, we introduced to the story that Rodrigo became popular and the 
leader of a gang asked him to join, and we asked participants to name pros and cons to joining 
or declining the offer. As the story evolved, we increased the severity of the crimes committed 
by Rodrigo to understand participants’ assessment of blame according to this evolution (from 
robbery to kidnapping to homicide). We then asked about blame for Rodrigo’s acts. Finally, we 
opened a discussion to see if any of the previous assessments would change if Rodrigo came from 
a higher socioeconomic background.
54
Initial conversations revolved around blame attribution to Rodrigo. Initially, the general sense was 
that Rodrigo is to blame for his decision to get involved in petty crime because there are 
alternatives. Arguments about Rodrigo’s responsibility pointed to the self-consciousness of one’s 
acts, the fact that being poor does not equal being mean, the fact that there are legal ways to 
obtain money in case of an emergency, such as borrowing and offering to do small tasks (e.g.,
washing clothes, sweeping) as repayment, and the opinion that crime is the easy way. Most 
participants from all groups shared these views. However, in most of the groups, positions that 
were more “understanding” of Rodrigo’s choices came up, particularly in the groups from CASA, 
La Tenda, and Supera (all in Northern Mexico). In groups where there was variation around 
individual blame, participants were more sensitive to Rodrigo’s socioeconomic background,
which was presented as a justification or explanation for his choices. Poverty and economic 
harshness attenuate responsibility. This view was clearly presented in the statements below:
“ … but in some cases, communities do not have the resources to give jobs. In my case, there 
are many people and young people who have to get involved in robbery because of need. This 
may be for children, for siblings, for the mother … ”
“I will obviously tell you this is wrong [to get involved in crime], but I have done it when I was 
younger. Now I do not do that anymore. When you look and look and look and cannot find [a 
job], the need calls you. If you have this option you will say yes. You have to ‘atorarle para sacar 
la papa’ [work to bring potatoes to the table].”
The fact that these youth are conscious about the wrongdoing involved in this decision may come 
from social desirability bias. These youth had already been in contact with the law and had taken
part in reintegration programs. Although many said they wouldn’t do it anymore because they 
understand the implications, it was clear that they still thought about the attenuating 
circumstances that explain why Rodrigo may have made a choice in the first place. 
When asked about joining a gang, the perception was similar. The general sense was that Rodrigo 
is to blame for his decision, and the main rationale presented by youth was that it was no longer 
out of necessity but out of greed—a choice to become a criminal. However, youth also pondered 
some circumstances that may explain this choice—one being that this might have not been a real 
choice as there might be retaliation from the gang. Again, in some cases, youth expressed an 
understanding of why Rodrigo may have made this decision. As put by one participant,
“We are saying no because we already lived through that, we know the difference between the 
good and bad. Knowing that you will make good money, imagine how that is like for a youth in 
need. Especially now in the pandemic. Le conviene. Claro que le conviene [it is in his interest to 
join, of course it is].”
It is also noteworthy that youth seemed to be conscious about the risk implied in the decision to
join a gang. Again, it is important to keep in mind that this came from youth who had participated 
in reintegration programs. As two participants expressed,
“With one saying I can tell you everything: don’t play with fire because you will get burned.”
55
“I want to share something a friend of mine posted. When you put a frog in water it will stay 
there, even if the water is getting hot. When the water is boiling, he cannot longer jump out. The 
message is that, sometimes, even if you know that the things you are doing are wrong, you feel 
comfortable and you want to stay there. Like Rodrigo, he might feel comfortable with the leader 
of the gang. But it may be that the water will boil and he will not be able to get out.”
As the story evolved, youth were asked about their perceptions as the severity of the crime 
increased. In collaboration with the gang, Rodrigo was involved in the kidnapping of a 
businessman. After the payment for his rescue did not come through, Rodrigo ended up killing 
the businessman following the gang leader’s order. Youth were asked about how much they 
blame Rodrigo for the homicide (using a scale from 1 to 7, with 7 being most to blame). Overall, 
the sense was that Rodrigo made his choice. While in all groups the majority of participants 
assigned 7 to Rodrigo being to blame for the homicide, in all CSO groups,
5 some participants 
assigned 4, 5, or 6, with the justification that Rodrigo was following orders and had no choice. 
Many blamed the leader equally or more than Rodrigo. As two participants expressed,
“I would give him a 4 because he was the weapon but was induced. It was not out of his own 
choice. If he was more conscious or mature he would have chosen differently. He is like a puppet.”
“It is not about whether you want it or not, ‘es a huevo’ [it is obligatory], they are telling you that 
you have to do it. If you say no, most likely they will tell you, ‘So then don’t do it and tambien the 
vamos a dar suelo’ [we will also kill you]. It is more about what you are mandated to do rather 
than wanting to do.”
Participants were then asked about how they attribute blame to Rodrigo, the government and 
society, and his mother for the circumstance in which Rodrigo ended up. While most participants, 
again, recognized the role of Rodrigo in making these choices and that bad choices led him to 
this scenario, society and the government were also blamed for the crimes committed by 
Rodrigo. A vast majority of participants from all groups blamed both. With regard to society, 
views are congruent with the findings about the mediators of normalization, as society is 
perceived as influencing people’s behavior. It is noteworthy that in several groups, the concept 
of society was intertwined with family and personal relations. Youth blamed the environment in 
which one develops. We found a tension between what is understandable and what is perceived 
as wrong. Youth may have recognized the wrongdoing involved in violence or its potential 
negative consequences, while at the same time showing empathy toward Rodrigo and 
understanding the circumstances that may have led him to make these choices. As mentioned 
above, we believe that this is a product of the incredibly hard work done by these youth to 
recognize and understand their choices while building toward a better future. It is a reflection of 
the realities they live. The three following statements are very telling in this respect:
“It is also society’s blame. I came to that point because of the environment in which I developed. 
It begins with society … I started getting along with certain people, to know certain things, I 
started doing things and that was it.”
5 In PILARES, there was no variation; all participants blamed Rodrigo. 
56
“I think sometimes it is not so much about society’s blame, but I think it is in some cases. We 
are all guilty, but no one is at the same time. We are not guilty in the sense that our parents 
taught us how to behave, maybe they taught us incorrectly. But they are not to blame because 
they were taught the same. Many times you have the decision and strength of choice, but if it's 
what you learned, what you were taught at home where you saw that it was normal to be involved 
in robbery and killings, maybe you made those choices when you were not mature enough to say,
‘This is not what I want, this is not right.’ … When you are not old enough to differentiate 
between what is right and what is wrong, you are not to blame. When you are mature enough,
then you can differentiate and you are to blame … ”
“I think everyone intervenes [in these decisions]. But also education comes from home. Rodrigo’s 
mother also had something to do. In my house, my mother is now more concerned about me, 
she calls me or gives me advice. I wake up and say, ‘a huevo’ [of course] I can. And she says I 
should not listen to people who give me bad advice.”
Regarding the government, most answers related to the lack of employment, social services, 
and opportunities for youth, as shown by one discussion among participants: 
Participant 1: “ … if the government were like the US government or like the government from 
another country, we wouldn’t be so fucked up.”
Participant 2: “It’s true. It has a lot to do with the areas where you live as well, how’s your school, 
how’re your surroundings. They just say, ‘Well, he just wanted to stay like that.’ You have to look 
deeper into the problem. Anyone can say, ‘He wanted to be like that,’ but she doesn’t know what 
that youth is going through.”
Participant 1: “I do feel it’s the government.”
Participant 2: “It’s what the youth bears on his back.”
Participant 3: “I think the government is the origin of this. As [other participant] says, these are 
the most abandoned areas, the ones getting least help from the government. As [Participant 2]
says, you don’t know what they bear on their back until you carry it. Sometimes you want to 
study, but there’s no benefit for it here, there’s no money.”
Participant 4: “Here the government … I have never had scholarships or anything like that. The 
only thing was a jacket, and that’s because they were campaigning.”
Finally, when asked if Rodrigo would be more to blame if he had more money, the general sense 
was that Rodrigo is still to blame because he made his own choices, dependent on his values 
However, again, for participants sensitive to the socioeconomic context (those mainly from the 
north), having sufficient money put more blame on Rodrigo, as he would only be acting for greed 
and there would not be a need that “justifies” his behavior. In the words of the participants,
“ … when they had nothing, they pushed him to do that. Apparently he had no option, but here, 
in this other situation, where he has money, it might be he has one. It might be that he wouldn’t 
have done it because, if he had the money, why did he want more? Only if he is greedy.”
Conversation:
Participant 1: “If he had money I would no longer say 5 because he didn’t have the need. He 
would be more guilty.”
Participant 2: “I also think that more, he made that out of his own choice.”
57
Participant 3: “If he had money, I think he would not have reacted this way. But because he did, 
he is more guilty. He was crazy.”
Participant 4: “When there is no need, you do it because ‘acá’ [referring to adrenaline that was 
being discussed or because you like it], but then when they tell you they will break you because 
you have been involved in crime, you get very scared because you know there was no need.”
Finally, we found that the type of victim is also relevant in shaping perceptions on crime. This 
occurs in two manners: through empathy and through risk assessment. Participants revealed 
empathy with certain victims. In all groups, most of them considered that the crime would be 
aggravated if the victim were an elderly worker lady. The elderly worker lady was seen as more 
vulnerable because of her age and her worker status, which indicate that she does not have much 
money. In a similar vein, there was empathy with the blue-collar worker because of his social 
class reflecting participants’ own backgrounds. From this perspective, the businessman and the 
politician were the preferred victims, as they carry the least moral load. Regarding the politician, 
it is interesting that a few participants even perceived the crimes against him as fair, given the 
corruption typically associated with politicians; one even mentioned an old Mexican saying: “a 
thief steals from a thief.”
On the other hand, this rationale was contradicted when it came to risk assessment. Participants 
perceived different risks associated with each victim. Most of them considered the acts against 
the businessman and the politician as being riskier because they would be more likely to have 
bodyguards and personal connections in the government to enforce punishment for the crime. 
Risks associated with the blue-collar worker had more to do with his physical strength, as many 
participants perceived that he would know how to fight and defend himself. Regarding the elderly 
lady, only one participant mentioned a risk associated with her, based on his personal experience, 
which is that she might die from the shock of being assaulted. From this perspective, the preferred 
victim was the elderly lady, which is completely opposite from an assessment based on empathy.
Protective Factors as Promoted by JPV Supported Interventions
The first phase of the performance evaluation had found pressing challenges remaining in the 
programmatic offer of the JPV activity in terms of addressing the underlying factors that cause 
violence in the first place and that interfere with program adherence and hamper social 
reintegration. These areas of opportunity were found mainly for interventions that are able to 
interrupt the community and family dynamics that produce antisocial behaviors, as well as 
interventions to address substance abuse. The first phase of the evaluation also found as an 
outstanding agenda the need to develop capacity among government agencies to generate a 
specialized offer of youth services. Relatedly, it identified the urgent need to strengthen promising 
models (mainly those that incorporate territorial strategies) and to scale them up to reach youth 
in need. The results gathered through the focus groups analysis build on these preliminary 
findings. 
Our results indicate that the studied programs do contribute to the development of protective 
factors in these youth. Some of them, mainly from Ciudad Juárez, expressed highly valuing CSO 
programs because of the otherwise nonexistence of social and cultural services with a youth 
perspective in the areas where they live. In general, youth consider that these programs help 
58
them to stay away from the dynamics that drive them to get involved with crime and violence. 
This was clearly stated by one participant from CASA, who used Rodrigo’s vignette to explain:
“ … to distract your mind from ‘Let’s go robbing.’ If [Rodrigo] were with CASA, he would be like, 
‘Let’s play football, let’s get a snack.’ Here, there’s always food. He would be more relaxed. ‘No, 
don’t go robbing, here we’ll give you a scholarship for your studies.’ So, he stops thinking about 
crime. The story would be different, no?”
Participants were generally positive about their experience with CSOs and identified clear 
benefits of education and employability, mental health, addiction treatment, and support 
networks. Nonetheless, challenges remain, and opportunity areas were identified in the group 
discussions. 
Education and employability
As mentioned above, most of the participants had dropped out of school and had not finished 
their upper-secondary education, with the exception of those in the PILARES group. CSOs and 
PILARES provide education services focused on helping youth find a job or undertake an 
economic activity of their own. The programs aim to support youth in continuing their formal 
education through distance-education systems, tutoring, and scholarships. In PILARES, education 
services integrate training in demanded skills in the labor market (e.g., Microsoft Office, English) 
and training in skills for self-employment (e.g., electricians, mechanics). CSOs are mainly centered 
on training for self-employment. 
After being involved in the programs, youth assign value to continuing with their studies and to 
learning new things, both for personal development reasons and to be occupied. They highly value 
the scholarships given to them by CSOs, as they do not think they would have been able to 
continue with their education otherwise. In most instances, education is perceived as a means to
access better employment opportunities, other than being a mason or working at the “maquila.” 
The most important challenge found relates to work training and labor intermediation. In some 
CSOs, youth stressed the need to enhance the relevance of training to help them find a job. This 
was particularly salient in Reinserta, where no work training is provided at all. Furthermore, 
CSOs’ capacity to find job vacancies is limited, if any in certain cases. Although some of them 
have direct relationships with private-sector organizations at the state level, like Consejo and 
Supera, and others, like Reinserta, have established direct agreements with restaurants and firms, 
these do not seem to be effective in supporting youth employability. In some cases, there are
profiling problems and youth do not last in these jobs. In other instances, jobs are too far away 
from where these youth live. In this context, self-employment might be an alternative, although 
no structured services for this purpose were identified beyond specific skills training. This is a 
highly sensitive issue given that, while we know that recidivism is not only responsive to income 
(Wilson et al., 2000) economic need may indeed drive recidivism. Moreover, these youth may 
face frustration after not being able achieve their goals and may have negative reactions that drive 
recidivism.
Mental health
CSOs provide both individual and group therapies. Participants mentioned that these have helped 
them to improve their emotional regulation and their family relationships. They also noted that 
59
they have changed their views in many ways, such as not wanting to be wandering and doing 
nothing anymore. Some of them highlighted that these are services too expensive for them to 
afford and that they would not have access to them otherwise. The exception is PILARES, where 
there are no psychological services, only group discussions and workshops for socioemotional 
skill development. Nonetheless, participants mentioned these as useful, as they can learn from 
others’ experiences.
We identified three main areas of opportunity, the first being staff profiles in CSO psychology 
departments. This was pointed out in one organization in particular where some participants 
considered there to be limited rapport from psychologists, as they come from backgrounds too 
distant from theirs and do not seem to fully understand the situations from which these youth 
come. The second area of opportunity is psychiatric treatment. Most of these youth had abused 
drugs and had been exposed to highly violent situations, either as perpetrators or victims. 
However, there was no indication made of services existent for these problems. The third area 
of opportunity is work with families. As mentioned above, domestic violence, substance abuse, 
and criminal activity are highly intertwined with family relations in this context. Although CSOs 
have incorporated family therapy in their service offerings, this is insufficient to address these 
complex family dynamics. 
Addiction treatment
All CSOs address addictions in group therapy; however, only a few (i.e., Supera, CASA, and La 
Tenda from the north) have put in place harm reduction programs. Based on participants’ 
comments, harm reduction programs seem to have some effectiveness in helping youth with drug 
abuse. This was particularly salient in discussions from La Tenda:
“I was slimmer than him, and now look. It helped me with consumption, but not only here, I went 
to Alcoholic Anonymous and it does help. When they encourage and advise you, you feel better.”
“I quitted drugs, I don’t care about them anymore, and I stopped seeing my friends from that 
time. I don’t talk to them anymore.” 
Nonetheless, addiction treatment remains a challenge and a sensitive issue to address, given the 
close relationship of substance abuse with violent and criminal behavior and considering that it 
affects the social spheres of most of these youth. As mentioned before, only a few CSOs have 
put in place specialized strategies in this area, and some of them only address it in group therapy; 
they do not admit youth who are active substance abusers. This is a major opportunity area for 
social reintegration programs. Promising intervention models from certain CSOs should be 
strengthened and shared with CSOs that are not fully addressing this issue yet. This should be 
reinforced with partnerships with specialized institutions to ensure that youth are assisted in this 
matter.
Support networks
A salient benefit of CSO programs, specifically of CASA, La Tenda, Consejo, and Supera, is that 
of support networks. Participants from these CSOs highlighted that they have made friends there, 
including both peers and staff. In PILARES, one participant also mentioned having generated a 
support network with women from a feminist discussion group. A relevant aspect in this regard 
is the follow-up done by these CSOs after program conclusion in inviting youth to regular 
60
activities to maintain relationships. To many, CSOs had become a safe place where they feel well. 
In the words of one participant from Consejo and CASA,
“ … my generation is one of the oldest. I've been here since it started. Then, I felt so happy to 
be here that I keep coming back, although it’s not the same anymore, I keep coming back because 
here’s where I found happiness.”
A remaining area of opportunity is strengthening the follow-up strategies of CSOs, as this 
component is not yet incorporated into every model. This is highly relevant to sustain programs' 
positive effects and to keep youth away from dynamics that drive them to violence and crime. 
After finishing the programs, there are very few opportunities for youth to engage in productive 
and health promoting activities, and the CSOs capacities are very limited to provide these services 
alone. 
61
VI. Conclusions and Policy Implications
Over the past decade and a half, Mexican youth have been raised amid an unprecedented wave 
of organized criminal violence. They comprise a particularly vulnerable population as targets of 
violence, but also as potential recruits. Criminal career paths start early in life, often through 
participation in gangs. The motivations driving youth involvement in criminal activity vary across 
contexts. In addition to economic opportunity and coercion, nonmaterial incentives and 
contextual factors play important roles in determining why some individuals embark on criminal 
life paths while others do not. This study contributes to our understanding of youth crime by 
investigating how they (i) perceive, (ii) rationalize, and (iii) condemn or justify acts of crime and 
violence. 
Our findings indicate that the gender, family and school environments, as well as social networks 
of youth significantly influence the normalization of certain acts of violence and their perceptions 
of gang involvement. Of particular interest is the role of neighborhood spaces for socialization as 
potential enablers of gang involvement and recruitment. Based on the survey evidence, we cannot 
disentangle the specific types of group activity, community organization, or social gathering that 
are correlated with gang involvement and recruitment. But the data indicate that risks of gang 
involvement increase among those affiliated with a generic category of community organization. 
It is very likely that community organizations in some neighborhoods have been infiltrated by 
members of criminal organizations. The focus groups shed light on how drug cartels’ dynamics 
(e.g., whether they have monopolistic control over the territory or not) and their relationships 
with the communities’ influence youth risk assessments of joining criminal organizations, 
indicating that prevention strategies should be tailored to the specific local crime dynamics. 
Hence, it is important that interventions aimed at preventing youth crime identify venues of 
socialization that minimize youth risks and that are insulated from criminal groups. School 
organizations seem to be one such avenue, as indicated in our analysis. 
Our findings also reveal that it is crucial to design interventions targeting young men with poor 
family relations and those who have been victims of domestic violence. In this regard, curbing the 
normalization of violence—particularly among men—seems to be critical to preventing youth 
criminal behavior. The survey analysis provides robust evidence indicating that men normalize 
violent behavior much more than women across all types of violence. From this perspective, 
families themselves should also be targets of intervention programs. Early interventions are 
central to preventing normalization, as they constitute the primary sphere of socialization. As 
found in the focus group discussions, violence is a form of social interaction that is transmitted 
from one generation to the next. Gender stereotypes and misogynist views are also learned at 
home and appear to be central in the normalization of violent behavior across all social spheres. 
Work in this area needs to be reinforced as part of the violence prevention agenda, as is work 
focused on substance abuse, which plays a major role in generating violence among relatives and 
affects all social spheres of youth at-risk. 
The evidence from our experiment on blame attribution reveals that, on average, youth in urban 
Mexico see lower-class peers who engage in criminal behavior as less deserving of blame 
compared to middle-class or more privileged individuals. In such cases, respondents were more 
likely to blame the government or society as a whole for the crimes committed. Respondents 
were also more likely to blame perpetrators who act as gang leaders than those who follow 
orders. The type of crime and the identity of the victim seem to be less influential and highly 
62
context-specific in shaping assessments of attribution of blame. Further research is needed to 
explore whether this apparent indifference with respect to the victim and the type of crime is 
due to increased normalization of violence or to a generalized stigmatization of all criminal 
behavior. The focus groups revealed some promising lines of research in this regard: empathy 
and risk assessments associated with the victim appear as relevant dimensions to be explored to 
further disentangle the mechanisms behind blame attribution. Taken together, these findings show 
that youth rationalize criminal behavior as the result of structural conditions and not necessarily 
due to individual agency. The focus group discussions also revealed the complexity underlying the 
rationalization processes of these youth. While most youth recognized the centrality of individual 
choices, society and the government were also blamed for crime and violence. This suggests that 
effective youth crime prevention also requires trust-building strategies to restore the social fabric 
and legitimize state institutions.
63
VII. Recommendations
1. Map local crime dynamics and enablers.
The aim should be to create a map of local criminal dynamics and potential enablers for 
recruitment, crime, and violence in high-violence areas. In particular, we consider it highly 
relevant to:
a. Create a map of gangs at the neighborhood level in collaboration with local 
police authorities to understand the potential reach and coverage of the territory, 
as well as their local crime dynamics, whether there are affiliations with organized 
crime, and the potential consequences such as gang recruitment. These efforts 
should be prioritized in high-violence areas.
b. Create a census of neighborhood or community organizations, including 
governmental organizations, NGOs, and informal organizations to identify the 
types of organization or group activity that are more likely to be correlated with 
patterns of gang activity, crime, and violence.
The evidence on how to prevent youth from joining gangs in the first place is very scarce. 
These efforts would be a first step towards understanding these dynamics and providing
data-driven evidence to design and target effective policy approaches. 
2. Expand the offer of youth prevention programs, addressing the mechanisms 
that drive violence normalization.
It is imperative to expand the services available to support at-risk youth, particularly 
focused on those at high risk for violence (secondary prevention) or to prevent recidivism 
of offenders in high violence areas. In this sense, we consider it urgent to:
a. Invest in the development of multidimensional therapy models, 
addiction treatment, and psychiatric models, specific for the Mexican 
context. The aim should be to enable the environment for specific and promising 
innovations to be able to thrive, to generate knowledge, and to identify new, 
effective solutions with the potential for scaling. Unfortunately, the evidence base 
on what works to reduce violence among youth is relatively narrow. 
Governments, CSOs, researchers, and the private sector are addressing this issue 
by conducting a variety of policy experimentations. The objective of aid 
programming should be on developing specific subject matter–related technical 
expertise, implementation, and monitoring-and-evaluation (M&E) capacities that 
allow strengthening, evaluating, and replicating models if effective. The evidence 
has suggested that these models are promising to prevent violence engagement 
and recidivism,6 and efforts should be implemented to enable innovation and 
adaptation around their principles. 
6 See “What Works to Prevent Violence Among Youth?”, USAID J-PAL–commissioned report, 2018. 
64
Existing programs, such as those supported by JPV, provide observational evidence 
(we do not yet have causal evidence) that some interventions components may 
work, for example, by providing support networks, harm reduction, or 
psychological support, among others. These experiences show that there is 
opportunity to successfully intervene, even before youth join gangs. Efforts and 
investments should be targeted to strengthen these interventions, to generate 
more evidence on whether and how they work, and to eventually scale up effective 
components.
b. Prioritize gender-based violence prevention interventions. A gender 
perspective can be undertaken with a cross-cutting approach in programs from 
both the government and CSOs. These can be reinforced with interventions to 
create awareness of sexual violence, gender stereotypes, and misogynist views.
c. Implement and test specific interventions aimed at supporting victims 
of violence, particularly the gender-based ones. For instance, work with 
peer networks for young women can be relevant to support them when facing 
intimate partner violence. Interventions at the parental level may also be relevant 
to raise awareness about the negative impacts of physical and verbal abuse, as well 
as discussions with youth exposed to domestic violence in which they can discuss 
and conscientize these negative impacts to help them understand, overcome, and 
prevent future occurrence.
3. Implement perspective-taking interventions. 
To inform participants about the drivers of youth involvement in criminal behavior and 
raise awareness about the risks and the impact—for them and for victims— associated 
with such behavior. Although these strategies may not be enough as stand-alone 
interventions for youth who have already been involved in crime, they may be helpful as 
a secondary prevention intervention. We propose the following:
a. Test police-youth or victim-youth dialogue programs that apply 
narrative-based approaches to increase empathy and trust, and define 
outcome measures that can be systematically evaluated.
b. Work with community organizations to assess mechanisms through 
which they may be enabling certain types of violence normalization and 
leverage their targeting (or capacity to reach at-risk youth) to 
implement perspective-taking interventions (see below), particularly 
focused on the risk of joining gangs. These interventions should consider variations 
in risk assessments resulting from specific local crime dynamics, mainly regarding 
drug cartels' types of activity.
4. Work with the private sector to expand employment opportunities for at-risk 
youth. Also, creating paid job-training programs and positions that are aligned with the 
labor market demand and train youths to fit those needs, improving profiling, and reducing 
rotation. For many of these youth, self-employment is a better alternative, as their profiles 
65
can hardly be aligned with available job vacancies. Thus, programs for entrepreneurship 
skill development should also be incorporated, adapted to youth characteristics.
5. Engage with government agencies to create and expand specialized capacity 
for the design and implementation of specific services aimed at preventing 
youth crime involvement and reducing recidivism. 
Increasing the capacity of CSOs to work with at-risk youth is paramount, especially given 
that the civil society sector has more flexibility to try out innovative approaches. 
However, there is also a need to generate specialized capacity within governments to 
support youth to build alternative pathways to crime and violence, which is of the greatest 
priority to achieve the needed reach and scale. In particular, we consider it urgent to 
accomplish the following:
a. Improve the offer of social services with a youth perspective in higher￾risk communities. There is a lack of mental health, education, employment, and 
cultural services tailored for this population in the places where they live. This 
means that both the design of programs and the service provision staff should be 
prepared to address the specific needs of these youth. This should be a 
governmental responsibility as it constitutes a condition of access to social rights 
for these youth.
b. Create programs that allow combining post-release support or 
alternatives to incarceration with comprehensive social services that are 
aimed at remedying the behavior leading to the original arrest and that promote 
safety, strength, and permanence in the community, schools, home, and place of 
employment while keeping youth accountable—this is the most promising 
approach to reduce recidivism sustainably. These services should coordinate 
physical and mental health services, education and employment support, and family 
counseling. There remain ample opportunities to engage with public institutions 
to generate this specific offer. 
6. Create a collaborative research model with academic researchers and 
practitioners to improve the accumulation of knowledge and inform evidence￾based policy. This could be done through the implementation of a cluster of coordinated 
research studies subjected to preregistration analysis, constant monitoring of results, and 
impact evaluations.
66
References
Achen, C. H., and Larry M. B. (2002). “Blind Retrospection: Electoral Responses to Drought, Flu, 
and Shark Attacks.” Presented at the annual meeting of the American Political Science 
Association, Boston.
Alleyne, E., Fernandes, I., & Pritchard, E. (2014). Denying humanness to victims: How gang 
members justify violent behavior. Group Processes & Intergroup Relations, 17(6), 750-762.
Arceneaux, K. (2003) “The Conditional Impact of Blame Attribution on the Relationship Between 
Economic Adversity and Turnout.” Political Research Quarterly 56(1): 67–75.
Azaola, E. (2016). Diagnóstico de las y los adolescentes que cometen delitos graves en México, 
UNICEF, Subsecretaría de Prevención y Participación Ciudadana de la Secretaría de Gobernación.
Ayala, M., & López, A. (2016, September 15). Prevenir delitos: crónica de una muerte anunciada. 
Animal Político. Retrieved from http://www.animalpolitico.com/blogueros-el-blog-de￾mexicoevalua/2016/09/15/prevencion-del-delito-cronica-una-muerte-anunciada/
Bartling, B., & Fischbacher, U. (2012). Shifting the blame: On delegation and responsibility. The 
Review of Economic Studies, 79(1), 67-87.
Bastian, B., Denson, T. F., & Haslam, N. (2013). The roles of dehumanization and moral outrage 
in retributive justice. PloS one, 8(4), e61842.
Becker, G. S. (1968). Crime and punishment: An economic approach. In The economic 
dimensions of crime (pp. 13-68). Palgrave Macmillan, London.
Becker, G. S. (1998). Accounting for tastes. Harvard University Press.
Bernasco, W., de Graaff, Th., Rouwendal, J., and Steenbeek, W. (2017), ‘Social Interactions and 
Crime Revisited: An Investigation Using Individual Offender Data in Dutch Neighborhoods’, 
Review of Economics and Statistics, 99, 622–36.
Billings, S. B., Deming, D. J., & Ross, S. L. (2019). Partners in crime. American Economic Journal: 
Applied Economics, 11(1), 126-50.
Bisbee, J. (2019). BARP: Improving Mister P Using Bayesian Additive Regression Trees. American 
Political Science Review, 113(4), 1060-1065.
Blumenthal, S., Gudjonsson, G., & Burns, J. (1999). Cognitive distortions and blame attribution in 
sex offenders against adults and children. Child abuse & neglect, 23(2), 129-143.
Blumstein, A. (Ed.). (1986). Criminal Careers and" Career Criminals," (Vol. 2). National 
Academies.
67
Bruce, D. (2007). “To be someone: Status insecurity and violence in South Africa.” In P. Burton 
(Ed.) Someone stole my smile, 57-68.
Calderón, L., Ferreira, O. R., & Shirk, D. A. (2018). Drug violence in Mexico: Data and analysis 
through 2017. Justice in Mexico, University of San Diego, April, 7.
Cárdenas, S. (2011) “Escuelas de doble turno en México. Una estimación de diferencias asociadas 
con su implementación,” Revista Mexicana de Investigación Educativa, 16(50), 801-827. 
Carrell, S. E., & Hoekstra, M. L. (2008). Externalities in the classroom: How domestic violence 
harms everyone's kids. American Economic Journal: Applied Economics.
Carvalho, L. S., & Soares, R. R. (2016). Living on the edge: Youth entry, career and exit in drug￾selling gangs. Journal of Economic Behavior & Organization, 121, 77-98.
Caudillo, M. and Torche, F. (2014). Exposure to Local Homicides and Early Educational 
Achievement in Mexico. Sociology of Education, 87(2):89-105.
Chapa Koloffon, L., & Ley, S. (2015). Prevención del Delito en México: ¿Cuáles son las 
prioridades? Mexico City: México Evalúa.
Centro de Investigación para el Desarrollo, A.C. Justicia para Adolescentes, 2016.
Chávez Villegas, C. (2020). Poverty, aspirations, and organized crime in Ciudad Juárez, Mexico. 
Victims & Offenders, 15(3), 330-349.
Cook, P. J., & Ludwig, J. (2005). Assigning deviant youths to minimize total harm (No. w11390). 
National Bureau of Economic Research.
Cozzarelli, C., Wilkinson, A. V., & Tagler, M. J. (2001). Attitudes toward the poor and attributions 
for poverty. Journal of social issues.
Crane, J. (1991). The epidemic theory of ghettos and neighborhood effects on dropping out and 
teenage childbearing. American journal of Sociology, 96(5), 1226-1259.
Damm, A. P., & Dustmann, C. (2014). Does growing up in a high crime neighborhood affect youth 
criminal behavior?. American Economic Review, 104(6), 1806-32.
Deming, D. (2011). Better schools, less crime. The Quarterly Journal of Economics, 126(4):2063–
2115.
Draca, M. and Machin, S. J. (2015). Crime and economic incentives. Annual Review of Economics, 
7:389–408. 
68
Ehrlich, I. (1973). Participation in illegitimate activities: A theoretical and empirical investigation. 
Journal of political Economy, 81(3), 521-565.
Ehrlich, I. (1973). Participation in illegitimate activities: A theoretical and empirical investigation. 
Journal of political Economy, 81(3), 521-565.
Eriksson, K. H., Hjalmarsson, R., Lindquist, M. J., and Sandberg, A. (2016). The Importance of 
Family Background and Neighborhood Effects as Determinants of Crime. Journal of Population 
Economics, 29(1), 219–62.
Felson, R. B., & Palmore, C. (2018). Biases in blaming victims of rape and other crime. Psychology 
of Violence, 8(3), 390.
Freedman, M. and Owens, E. G. (2014). Your friends and neighbors: Localized economic 
development and criminal activity. Review of Economics and Statistics.
García-Ponce, Omar, Lauren Young, and Thomas Zeitzoff. Forthcoming. Anger and Support for 
Retribution in Mexico’s Drug War. Journal of Peace Research. 
Gaviria, A., & Raphael, S. (2001). School-based peer effects and juvenile behavior. Review of 
Economics and Statistics, 83(2), 257-268.
Ghetti, S., & Redlich, A. D. (2001). Reactions to youth crime: Perceptions of accountability and 
competency. Behavioral sciences & the law, 19(1), 33-52.
Gibson, J. L., & Gouws, A. (1999). Truth and reconciliation in South Africa: Attributions of blame 
and the struggle over apartheid. American Political Science Review, 93(3), 501-517.
Glaeser, E. L., Sacerdote, B., and Scheinkman, J. A. (1996). Crime and social interactions. The 
Quarterly Journal of Economics, 111(2):507–548.
Gudjonsson, G. H., & Bownes, I. (1991). The attribution of blame and type of crime committed: 
data for Northern Ireland. The Journal of Forensic Psychiatry, 2(3), 337-341.
Gudjonsson, G. H. (1989). Compliance in an interrogative situation: A new scale. Personality and 
Individual differences, 10(5), 535-540.
Hamilton, V. L., & Hagiwara, S. (1992). Roles, responsibility, and accounts across cultures. 
International Journal of Psychology, 27(2), 157-179.
Haynie, D. L. (2001), Delinquent Peers Revisited: Does Network Structure Matter?. American 
Journal of Sociology, 106(4), 1013–57.
Hjalmarsson, R., and Lindquist, M. J. (2012), Like Godfather, Like Son: Exploring the 
Intergenerational Nature of Crime. Journal of Human Resources, 47, 550–82.
69
Hjalmarsson, R., and Lindquist, M. J. (2013), The Origins of Intergenerational Associations in 
Crime: Lessons from Swedish adoption data. Labour Economics, 20, 68–81.
Jaitman, L., & Ajzenman, N. (2016). Crime concentration and hot spot dynamics in Latin America 
(No. IDB-WP-699). IDB Working Paper Series.
Jones, N. P. (2013). Understanding and addressing youth in “gangs” in Mexico. Washington, 
Woodrow Wilson International Center for Scholars, Working Paper Series on Civic Engagement and Public 
Security in Mexico.
Kluegel, James R., and Elliot R. Smith. 1986. Beliefs about Inequality: Americans’ Views of What 
Is and What Ought to Be. New York: Aldine De Gruyter.
Le Cour, R. (2018). Rediseñar la estrategia de seguridad pública desde lo local. In Léase (si 
quiere gobernar en serio. México Evalúa.
Lerner, Jennifer S, Julie H Goldberg and Philip E Tetlock. 1998. “Sober second thought: The effects of 
accountability, anger, and authoritarianism on attributions of responsibility.” Personality and Social 
Psychology Bulletin 24(6):563–574. 
Levin, I., Sinclair, J. A., & Alvarez, R. M. (2016). Participation in the wake of adversity: Blame 
attribution and policy-oriented evaluations. Political Behavior, 38(1), 203-228.
Levitt, S. D., & Lochner, L. (2001). The determinants of juvenile crime. In Risky behavior among 
youths: An economic analysis (pp. 327-374). University of Chicago Press.
Lindquist, M. J., & Zenou, Y. (2019). Crime and networks: ten policy lessons. Oxford Review of 
Economic Policy, 35(4), 746-771.
Lochner, L. and Moretti, E. (2004). The effect of education on crime: Evidence from prison 
inmates, arrests, and self-reports. American Economic Review, 94(1):155–189.
Ludwig, J., Duncan, G. J., and Hirschfield, P. (2001), Urban Poverty and Juvenile Crime: Evidence 
from a Randomized Housing-Mobility Experiment. Quarterly Journal of Economics, 116(2), 655–
79.
Malhotra, N., & Kuo, A. G. (2008). Attributing blame: The public's response to Hurricane 
Katrina. The Journal of Politics, 70(1), 120-135.
McBride, D. C., VanderWaal, C. J., Terry, Y. M., & VanBuren, H. (1999). Breaking the cycle of 
drug use among juvenile offenders. Washington, DC: National Institute of Justice.
McGraw, K. M. (1990). Avoiding blame: An experimental investigation of political excuses and 
justifications. British Journal of Political Science, 20(1), 119-131.
70
McGraw, K. M. (1991). Managing blame: An experimental test of the effects of political accounts. 
The American Political Science Review, 1133-1157.
Michaelsen, M. M., & Salardi, P. (2020). Violence, psychological stress and educational 
performance during the “war on drugs” in Mexico. Journal of Development Economics, 143, 
102387.
Mitchell, O., Wilson, D. B., & MacKenzie, D. L. (2007). Does incarceration-based drug treatment 
reduce recidivism? A meta-analytic synthesis of the research. Journal of Experimental 
Criminology, 3(4), 353–375.
Moncada, E. (2020). The politics of criminal victimization: Pursuing and resisting power. 
Perspectives on Politics, 18(3), 706-721.
Montgomery, J. M., & Olivella, S. (2018). Tree-Based Models for Political Science Data. American 
Journal of Political Science, 62(3), 729-744.
Peffley, Mark. (1984). The Voter as Juror: Attributing Responsibility for Economic Conditions. 
Political Behavior 6(3): 275–94.
Pettit, B., & Western, B. (2004). Mass imprisonment and the life course: Race and class inequality 
in US incarceration. American sociological review, 69(2), 151-169.
Raphael, S., & Sills, M. (2006). Urban crime, race, and the criminal justice system in the United 
States. A Companion to Urban Economics, 515-535.
Rogers, P., Josey, N., & Davies, M. (2007). Victim age, attractiveness and, abuse history as factors 
in the perception of a hypothetical child sexual abuse case. Journal of Sexual Aggression, 13(2), 
121-137.
Rogers, P., Wczasek, R., & Davies, M. (2011). Attributions of blame in a hypothetical internet 
solicitation case: Roles of victim naivety, parental neglect and respondent gender. Journal of 
Sexual Aggression, 17(2), 196-214.
Rothstein, R. (2004). Class and schools. Teachers College, Columbia University.
Sampson, R. J., & Laub, J. H. (2005). A life-course view of the development of crime. The Annals 
of the American Academy of Political and Social Science, 602(1), 12-45.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A 
multilevel study of collective efficacy. science, 277(5328), 918-924.
Shaver, K. G. (2012). The attribution of blame: Causality, responsibility, and blameworthiness. 
Springer Science & Business Media.
71
Sirotkina, E., & Zavadskaya, M. (2020). When the party’s over: political blame attribution under 
an electoral authoritarian regime. Post-Soviet Affairs, 36(1), 37-60.
Stevens, M. (2018). Preventing at-risk children from developing antisocial and criminal behaviour: 
a longitudinal study examining the role of parenting, community and societal factors in middle 
childhood. BMC psychology, 6(1), 1-12.
Sviatschi, M. M. (2018). Making a narco: Childhood exposure to illegal labor markets and criminal 
life paths. Manuscript, Department of Economics, Princeton University. https://rpds. princeton. 
edu/sites/rpds/files/sviatschi making-a-narco march2018. Pdf
UN Office on Drugs and Crime (UNODC), United Nations Office on Drugs and Crime data. 
UNODC, 17 2020, available at: dataunodc.un.org [accessed 13 November 2020] 
Warr, M. (2002), Companions in Crime: The Social Aspects of Criminal Conduct, Cambridge, 
Cambridge University Press.
Weisburd, D. (2015). The law of crime concentration and the criminology of place. Criminology, 
53(2), 133-157.
72
Appendix
Section A. Supporting Information - Data and Methodology 
In the case of Mexico, homicide data are available from police investigations, compiled by the 
Executive Secretariat of the National System of Public Security (SESNSP), and from death 
certificates, compiled by the National Institute of Geography and Statistics (INEGI). While INEGI 
data tend to be more precise, their disadvantage is that they are usually delayed for more than a 
year. For this reason, we relied on SESNSP data, which covered the year we examined (2020) at 
the municipal level. We only examined intentional homicides.
Although homicide statistics are considered quite accurate, they do not capture the full extent 
of violence within communities. They do not give us a grasp on the extent of nonhomicidal crime
or on residents’ perceived level of crime and insecurity in their areas (irrespective of reality). 
Both of these factors likely influence residents’ behavior as well and are relevant to our analysis. 
Thus, we incorporated these measures in our sampling design (Methods 2 and 3). However, this 
information is not typically available at the municipal level. National crime and victimization 
surveys, such as the ENSU and the National Survey on Victimization and Public Security 
Perception (ENVIPE), often collect this information with sampling strategies for higher-level 
geographic units (e.g., metropolitan area or city). Because of this, these statistics are not 
representative at the municipal level.
To circumvent this issue and ensure stratification at a more fine-grained level of the municipality, 
we relied on innovative methods to extrapolate these data from such surveys and create 
municipal-level estimates. The method we utilized is called MRP, a form of small-area estimation, 
which is used widely by researchers to predict estimates of certain variables at geographic units 
below a survey’s level of representativity (e.g., see the works of Park et al. [2006], Hanretty 
[2019], and Butz and Kehrberg [2016]). Using MRP, we relied upon survey questions regarding 
crime from the most recent ENSU survey and the 2015 intercensus to generate our municipal 
estimates. In particular, we utilized the following two questions from ENSU for our estimates:
1. PERCEIVED SECURITY: In terms of crime, do you consider that to live in (CITY) 
currently is … [safe, unsafe]? En términos de la delincuencia, ¿considera que vivir 
actualmente en (CIUDAD), es … [seguro, inseguro]?
2. VICTIMIZATION: During the past year [insert year], that is to say from January to 
today, has a member of your household (including yourself) been victim of (INSERT 
TYPE OF CRIME) on card A? Durante este año [insert year], es decir, de enero a la fecha, 
¿algún integrante de este hogar incluido usted, sufrieron la situación (CÓDIGO DE 
INCIDENCIA) de la tarjeta A?
1. Robbery or assault in the street or in public transportation? Yes or No. Robo o asalto en 
la calle o en el transporte público (incluye robo en banco o cajero automático)? Sí o No?
2. Threats, pressure, or deception to demand money or goods or to do something/not to 
do something (extortion, blackmail)? Yes or No? Amenazas, presiones, o engaños para 
exigir dinero o bienes; o para que hiciera algo o dejara de hacerlo (extorsión)? Sí o No?
73
For the first measure, perceived security, we coded the variable as 1 if an individual reported 
feeling “insecure” and 0 if an individual reported feeling “secure” in their city. For our second 
measure, victimization, we coded the variable as 1 if the individual reported that a member of 
their household had either experienced robbery or extortion in the past year and coded 0 if not.
With these questions, we then estimated the relationship between various individual-level 
characteristics and their responses on these selected survey questions. We did this through 
multilevel regression, where we determined the relationship between selected characteristics—
in this case age, gender, education, and occupation—and reported (i) insecurity and (ii) household 
victimization. This involved two separate regression models, one for each outcome variable. The 
regression also factored in geographic location, with individuals’ municipalities (unrepresentative 
unit of interest) nested within their states (geographic unit of the survey).
Once these regression estimates were calculated, we then post-stratified them. This involved 
weighting our estimates by the prevalence of each type of individual within each municipality 
based on their individual-level characteristics (i.e., age, education, etc.). This “prevalence” was 
calculated by determining the population of each type of individual within a municipality according 
to the 2015 intercensus. The regression estimates, weighted in this manner, generated a 
municipal-level estimate (one for each municipality) for the most likely response to each of the 
two survey questions.
Through this process, we calculated two municipal-level estimates: one quantifying nonhomicidal 
victimization and the other quantifying residents’ perception of community security in their 
municipality. We calculated these values for all municipalities included in the most recent ENSU 
survey, yielding estimates for 157 municipalities. These values can be interpreted as a ranking of 
perceived security and victimization among the municipalities we examined. They allowed us to 
determine within our sample of municipalities how they rank comparatively in terms of these two 
values. However, these measures do have error associated with them and cannot be compared 
to estimates outside of these analyses. This error was predominantly created by limitations due 
to question wording and the need to match variables between the census and survey. We had to 
match individual-level responses on the ENSU survey to biographical information about the head
of household, as certain attributes were only recorded at the head-of-household level (e.g., 
education level, occupation) in this survey. We used these head-of-household characteristics 
when considering the prevalence of each type of individual in the intercensus. Thus, these 
measures are not perfect individual-level measures, but did provide us with comparable 
intersample estimates. For this reason, we interpreted them as a ranking.
We confirmed the adequacy of our estimates and explored the correlation between these 
variables with a simple correlation analysis, shown in Figure 1. 
Figure 1. Correlations between MRP estimates and true values (means)
74
From Figure 1, we can confirm that our MRP estimates and the true values are highly correlated 
(for those municipalities for which we had a representative estimator). We also see a correlation 
between homicide rates and MRP insecurity perception estimates, but a nearly zero correlation 
between homicide rates and MRP victimization estimates. We observe a high correlation between 
MRP security and victimization estimates.
To further explore these results, we ran mixed-effects models with random intercepts by state. 
Results are shown in Table 1, on which population, homicide, and homicide rate are rescaled as 
values ranging from 0 to 1.
Table 1. Mixed-Effects Models with Random Intercepts
For the security measures on Table 1, as the population increases, so does insecurity. As the 
homicide rate increases, so does the insecurity index. Total homicide does not have a significant 
relationship with this index. In terms of the victimization index, there is little relationship with 
the included variables (the coefficient sizes are very small and confidence intervals cross zero). 
As population increases, the victimization index increases slightly. However, total homicide and 
homicide rate do not have significant relationships with victimization index, as confidence 
intervals include zero in both cases. From this analysis, we observe that in urban Mexico, there 
are places with homicide violence but not much other violence and vice versa, and there does 
75
seem to be little correlation between homicide rates and victimization estimates. There are 
correlations between homicide rates and security perceptions, although these are not high. 
Therefore, we divided the sampling units (using the ENSU sample) by levels of homicide, 
nonhomicide violence (victimization), and insecurity perception.
The survey was applied to a random selection of respondents using Mexico’s electoral sections 
as the primary sampling units (PSUs). PSUs were selected for the sample through a systematic 
method of stratified probability proportional to size. Each PSU in the sampling frame was assigned 
to a nonoverlapping sample stratum based on three variables mentioned above: homicide rate,
reported nonhomicidal crime, and perceived insecurity. After analyzing both explicit and implicit 
sampling strategies, we decided to use an implicit strategy because the resulting sample was more 
representative of our target population in terms of community security.
76
Section B. Survey Questionnaire 
Hola, mi nombre es ______. Estoy haciendo entrevistas para un estudio sobre eventos actuales en México para la Universidad de 
George Washington. Tu participación nos ayudará entender mejor algunos asuntos del país. Todo lo que comentes será 
completamente anónimo y la participación es voluntaria. En cualquier momento podemos parar la entrevista o saltarnos preguntas 
si así lo deseas.
Si tienes cualquier pregunta después, puedes contactar a las personas en esta tarjeta [ENTREGAR AL ENTREVISTADO LA 
TARJETA DE INFORMACIÓN].
¿Tienes alguna pregunta? [SI SÍ, RESPONDE LAS PREGUNTAS DEL ENTREVISTADO. SI NO, CONTINÚA].
BLOQUE 1
P1. ¿Cuántos años tienes?
____________ 
99) NS/NC 
P2. ¿Creciste en este municipio? [SI RESPONDE NO, PREGUNTAR Y ANOTAR EN QUÉ MUNICIPIO Y ESTADO 
CRECIÓ O PASÓ LA MAYOR PARTE DE SU INFANCIA]
1) Si, crecí acá
2) No, crecí en: Municipio:________ Estado:___________
98) NS
99) NC 
BLOQUE 2
Ahora te voy a leer una serie de preguntas para las cuales no hay respuestas correctas o incorrectas, solo opiniones.
P3. [MOSTRAR TARJETA] En tu opinión, ¿cuál de los siguientes grupos crees que sería más efectivo para 
mantener el orden y la seguridad en tu comunidad?
[ESPERAR RESPUESTA]
1) Ejército
2) Marina
3) Policía federal / guardia nacional
4) Policía estatal
5) Policía municipal
6) Un grupo de autodefensa
7) El crimen organizado
8) Otros
98) NS
99) NC
P4. [MOSTRAR TARJETA] En una escala de 1 a 7, donde 1 significa que NO CONFÍAS NADA y 7 significa que 
CONFÍAS MUCHO, ¿qué tanto confías en ___ [INSERTAR ALEATORIAMENTE]? 1 Nada
2
3
4
5
6
7 Mucho
NS/NC
Tus vecinos 1 2 3 4 5 6 7 99
Tu familia 1 2 3 4 5 6 7 99
La policía municipal 1 2 3 4 5 6 7 99
77
La policía estatal 1 2 3 4 5 6 7 99
El ejército 1 2 3 4 5 6 7 99
La marina 1 2 3 4 5 6 7 99
Policía federal / guardia nacional 1 2 3 4 5 6 7 99
El gobierno federal 1 2 3 4 5 6 7 99
El gobierno estatal 1 2 3 4 5 6 7 99
El gobierno municipal 1 2 3 4 5 6 7 99
BLOQUE 3
P5. [MOSTRAR TARJETA] En una escala de 1 a 7, donde 1 es muy inseguro y 7 es muy seguro,¿qué tan seguro 
crees que es vivir en [INSERTAR NOMBRE DE MUNICIPIO] actualmente? 
[ESPERAR RESPUESTA]
1) 1 Muy inseguro
2) 2
3) 3
4) 4
5) 5
6) 6
7) 7 Muy seguro
99) NS/NC
P6. ¿Con qué frecuencia crees que en esta colonia alguien como tú haya pasado por alguna de las siguientes 
situaciones en los últimos 12 meses? ¿Nunca, rara vez, de vez en cuando, o siempre? Nunca Rara vez De vez en cuando Siempre
NS/NC
Ha sentido miedo de salir a la calle por temor a su seguridad personal 1 2 3 4 99
Ha pagado por protección personal 1 2 3 4 99
Ha presenciado o visto un homicidio 1 2 3 4 99
Ha visto personas con armas que no son policías o militares 1 2 3 4 99
P7. En los últimos 12 meses, has sido o no has sido víctima de ____?
Sí No NS NC
Robo de casa o negocio 1 2 98 99
Robo de auto 1 2 98 99
Asalto en transporte público 1 2 98 99
Herido/a por arma 1 2 98 99
Tentativa de homicidio 1 2 98 99
Extorsión 1 2 98 99
Secuestro 1 2 98 99
Abuso sexual 1 2 98 99
P8. En los últimos 12 meses alguna de las personas que viven en tu hogar ha sido o no ha sido víctima de ____?
Sí No NS NC
Robo de casa o negocio 1 2 98 99
Robo de auto 1 2 98 99
Asalto en transporte público 1 2 98 99
78
Herido por arma 1 2 98 99
Tentativa de homicidio 1 2 98 99
Homicidio 1 2 98 99
Extorsión 1 2 98 99
Secuestro 1 2 98 99
Abuso sexual 1 2 98 99
BLOQUE 4
Ahora te voy a contar una historia de una situación que pasa frecuentemente en muchos lugares de nuestro país. Cuando 
termine de leer, te pediré tu opinión sobre la situación.
Rodrigo es un joven mexicano que creció en una familia [DE BAJOS RECURSOS CON MUCHAS CARENCIAS/DE CLASE 
MEDIA DONDE NO LE FALTÓ NADA].
Desde muy temprana edad empezó a delinquir y recientemente fue arrestado por [ROBARLE EL CELULAR / EXTORSIONAR 
/ SECUESTRAR / ASESINAR] A [UN OBRERO/ UN EMPRESARIO / UN POLÍTICO LOCAL]. 
Rodrigo confesó a las autoridades haber cometido este crimen [Y SER EL LÍDER DE UNA BANDA CRIMINAL / Y DIJO 
QUE ESTABA SIGUIENDO ÓRDENES DEL LÍDER DE UNA BANDA CRIMINAL].
P9. Primero me gustaría que pienses qué tan culpable consideras que es Rodrigo por lo que sucedió en esta 
historia. En una escala del 1 al 7 donde 1 significa que no es culpable y 7 significa que es completamente culpable, 
¿qué tan culpable crees que es Rodrigo? 
[MOSTRAR TARJETA 4]
1) 1 No es culpable
2) 2
3) 3
4) 4
5) 5
6) 6
7) 7 Completamente culpable
98) NS
99) NC
P10. Algunas personas piensan que el gobierno y la sociedad son en parte responsables por lo que pasó en esta 
historia. Otros consideran que no tienen ninguna responsabilidad. En una escala del 1 al 7 donde 1 significa nada de 
responsabilidad y 7 mucha responsabilidad, ¿cuánta responsabilidad crees que tiene____? [MOSTRAR TARJETA 5] 1 Ninguna responsabilidad
2 3 4 5 6
7 Mucha 
responsabilidad
NS
NC
El Gobierno 1 2 3 4 5 6 7 98 99
La sociedad 1 2 3 4 5 6 7 98 99
79
P11. Ahora, te pediré que pienses en las posibles razones o causas por las que Rodrigo cometió este crimen. Por 
favor menciona las tres primeras razones que te vienen a la mente. 
[RESPUESTA ESPONTÁNEA, ESCRIBIR RAZONES POR SEPARADO]
RAZÓN A __________________________
RAZÓN B __________________________
RAZÓN C __________________________
98) NS
99) NC
P12. ¿Cómo consideras que Rodrigo debe ser castigado? 
[LEER OPCIONES 1 A 3]
1) Debe ser arrestado y procesado por autoridades
2) Debe ser castigado por otro grupo criminal 
3) Debe ser sujeto de castigo por parte de la comunidad o un grupo de autodefensa
4) No debe ser castigado [NO LEER ESTA OPCIÓN]
5) Otro [NO LEER ESTA OPCIÓN]
98) NS
99) NC
P13. Pensando en tu colonia, ¿qué tan probable es que algo como lo que ocurrió en la historia de Rodrigo pase 
aquí?
[LEER OPCIONES 1 A 4]
1) Muy probable
2) Algo probable
3) Poco probable
4) Nada probable
98) NS
99) NC
P14. En una escala del 1 al 7, donde 1 es nada violenta y 7 es muy violenta, ¿qué tan violentas son las siguientes 
acciones? [MOSTRAR TARJETA]
[ROTAR]
1 Nada
2
3
4
5
6
7 Muy
NS
NC
Tener sexo con alguien sin su consentimiento 1 2 3 4 5 6 7 98 99
Decirle piropos a alguien 1 2 3 4 5 6 7 98 99
Criticar a alguien por cómo se viste 1 2 3 4 5 6 7 98 99
Besar a alguien sin pedirle permiso 1 2 3 4 5 6 7 98 99
Golpear a alguien 1 2 3 4 5 6 7 98 99
Saludar en la calle a alguien desconocido 1 2 3 4 5 6 7 98 99
Vender drogas 1 2 3 4 5 6 7 98 99
Burlarse de alguien 1 2 3 4 5 6 7 98 99
Ayudarle a un desconocido 1 2 3 4 5 6 7 98 99
BLOQUE 5
P15. Ahora por favor dime si estás de acuerdo o en desacuerdo con las siguientes afirmaciones 
[ES PERAR RESPUESTA Y PREGUNTAR MUY O ALGO]:
80
Muy de 
acuerdo
Algo de 
acuerdo
Ni de 
acuerdo ni 
desacuerdo Algo en [ESPONTÁ desacuerdo NEA]
Muy en 
desacuerdo
NS
NC
Si te sumas a una banda criminal estás más seguro, y tienes protección 1 2 3 4 5 98 99
Puede ser divertido estar en una banda criminal 1 2 3 4 5 98 99
La mayoría de quienes forman parte de una banda criminal terminan 
asesinados 1 2 3 4 5 98 99
La mayoría de quienes forman parte de una banda criminal terminan 
heridos 1 2 3 4 5 98 99
La mayoría de quienes forman parte de una banda criminal terminan en la 
cárcel 1 2 3 4 5 98 99
P16. ¿Conoces en persona a alguien que pertenezca o haya pertenecido a una banda criminal?
[ESPERAR RESPUESTA]
1) Sí
2) No
98) NS
99) NC
RECUERDA QUE SI ALGUNA PREGUNTA TE HACE SENTIR INCÓMODO(A) NO TIENES QUE CONTESTAR
P17. Durante tu infancia, ¿qué tan común era el abuso verbal, insultos fuertes, u ofensas hirientes, en tu casa?
[LEER OPCIONES 1 a 4]
1) Muy común
2) Algo común
3) Poco común
4) No era común
99) NS/NC
P18. Durante tu infancia, ¿qué tan comunes eran los golpes en tu casa?
[LEER OPCIONES 1 a 4]
1) Muy común
2) Algo común
3) Poco común
4) No era común
99) NS/NC
P19. Durante tu vida de estudiante, en cualquier escuela a la que asististe, algún maestro o personal de escuela, o 
tus compañeros:
Sí No NS NC
¿Te golpearon? 1 2 98 99
¿Te atacaron con un cuchillo o un arma? 1 2 98 99
¿Te hicieron cyberbullying o chantajearon por medio de redes sociales / 
celulares / correo electrónico? 1 2 98 99
¿Destruyeron tus pertenencias? 1 2 98 99
¿Te robaron? 1 2 98 99
¿Te tocaron sin tu consentimiento? 1 2 98 99
¿Te pidieron favores sexuales? 1 2 98 99
81
BLOQUE 6
P20. ¿Actualmente, estás afiliado a alguna organización que aparece en esta tarjeta? 
[MOSTRAR TARJETA Y MARCAR TODAS LAS QUE INDIQUE EL ENTREVISTADO]
Sí No NS NC
Organización religiosa 1 2 98 99
Equipo de deporte 1 2 98 99
Organización de la colonia/barrio 1 2 98 99
Partido político 1 2 98 99
Grupo musical, de artes o danza 1 2 98 99
Organización estudiantil 1 2 98 99
P21. ¿Has participado alguna vez en algún programa social para jóvenes que aparece en esta tarjeta? 
[MOSTRAR TARJETA Y MARCAR TODAS LAS QUE INDIQUE EL ENTREVISTADO]
Sí No NS NC
Prevención de la violencia 1 2 98 99
Prevención o reducción de consumo de drogas 1 2 98 99
Empleo juvenil 1 2 98 99
Deportes 1 2 98 99
Participación comunitaria 1 2 98 99
Arte o cultura 1 2 98 99
P22. ¿Qué tan satisfecho estuviste con ese o esos programas?
[ESPERAR RESPUESTA]
1) Muy satisfecho
2) Satisfecho
3) Neutral
4) Insatisfecho
5) Muy insatisfecho
98) NS
99) NC
P23. ¿Te interesaría o no te interesaría participar en un programa de ___? 
Sí No NS NC
de empleo (bolsa de trabajo, emprendimiento, elaboración de CV) 1 2 8 9
de capacitación para el trabajo 1 2 8 9
religioso o espiritual 1 2 8 9
de arte/cultura 1 2 8 9
de deporte 1 2 8 9
de prevención de la violencia 1 2 8 9
educativo 1 2 8 9
P24. ¿Crees que existen opciones disponibles en tu comunidad si quieres formar parte de un programa social 
juvenil?
[ESPERAR RESPUESTA]
1) Sí
2) No
98) NS
99) NC 
82
BLOQUE 7
Dado que el Covid-19 ha impactado nuestras vidas, nos gustaría hacerte una pregunta relacionada con ello.
P25. Durante la pandemia, muchas familias se han visto afectadas en nuestro país y en el mundo. ¿Podrías decirme 
si desafortunadamente alguno de tus familiares o amistades cercanas falleció a causa de Covid-19?
1) Sí
2) No
98) NS
99) NC
BLOQUE 8
Muchas gracias por contestar todas estas preguntas. Ya nos faltan muy pocas para terminar.
P26. ¿Hasta qué año o grado _____ en la escuela?
A. Aprobaste
B. Aprobó tu papá
C. Aprobó tu mamá
1) Ninguno 
2) Preescolar 
3) Primaria 
4) Secundaria 
5) Carrera técnica con secundaria terminada 
6) Normal básica (con antecedente en secundaria) 
7) Preparatoria o bachillerato 
8) Carrera técnica con preparatoria terminada 
9) Licenciatura o profesional 
10) Maestría o doctorado 
98) NS
99) NC
P27. ¿Quién es el principal proveedor o proveedora en tu hogar?
1) Madre
2) Padre
3) Ambos padres
4) Abuelo(s)
5) Hermano(s)
6) Tío(s)
7) Pareja
8) Otro
98) NS
99) NC
P28. ¿La semana pasada _____?
A. Usted
B. El principal proveedor del hogar
1) Trabajó
2) Tenía trabajo, pero no trabajó
3) Buscó trabajo
83
4) Es estudiante
5) Se dedica a los quehaceres del hogar 
6) Es jubilado(a) o pensionado(a)
7) Está incapacitado(a) permanentemente para trabajar 
8) No trabajó 
99) NR
SÓLO SI RESPONDIÓ CÓDIGO 1 O 2 EN RESPUESTA ANTERIOR
P29. En su trabajo o negocio de la semana pasada _____ fue ...
A. Usted
B. El principal proveedor del hogar
1) jornalero(a) o peón?
2) empleado(a) u obrero(a)? 
3) trabajador(a) por su cuenta?(no contrata trabajadores) 
4) patrón(a) o empleador(a)? (contrata trabajadores) 
5) trabajador(a) sin pago? 
99) NR
P30. Por favor dime si tu o cualquier miembro de este hogar tienen acceso a los siguientes servicios en casa 
ACTUALMENTE. También me gustaría saber si también tenían acceso a dichos servicios CUANDO TU TENÍAS 10 
años.
Sí No NS/NC
¿Automóvil actualmente? 1 2 99
¿Automóvil cuando tenías 10 años? 1 2 99
¿Lavadora actualmente? 1 2 99
¿Lavadora cuando tenías 10 años? 1 2 99
¿Drenaje dentro de la casa actualmente? 1 2 99
¿Drenaje dentro de la casa cuando tenías 10 años? 1 2 99
¿Computadora actualmente? 1 2 99
¿Computadora cuando tenías 10 años? 1 2 99
¿Internet actualmente? 1 2 99
¿Internet cuando tenías 10 años? 1 2 99
¿Teléfono celular actualmente? 1 2 99
¿Teléfono celular cuando tenías 10 años? 1 2 99
¿Servicio doméstico actualmente? 1 2 99
¿Servicio doméstico cuando tenías 10 años? 1 2 99
P31. El pasado 6 de junio se llevaron a cabo las elecciones para elegir Diputados Federales. Algunas personas no 
pudieron ir a votar por falta de tiempo, porque estaban ocupados, porque tenían miedo de contagiarse, o porque 
simplemente querían expresar su descontento con los partidos políticos. ¿Tú fuiste a votar o no fuiste a votar en la 
elección del 6 de junio?
[ESPERAR RESPUESTA]
1) Sí
2) No
99) NC
P32. [SÓLO SI RESPONDIÓ CÓDIGO 1 EN PREGUNTA ANTERIOR]
[ENCUESTADOR: MUESTRE LA TARJETA Y DESPUÉS HAGA LA PREGUNTA]
Te voy a entregar esta tarjeta donde vienen los partidos que compitieron en la elección de Diputados Federales. ¿Por cuál 
partido votaste?
1) Partido Acción Nacional (PAN)
84
2) Partido Revolucionario Institucional (PRI)
3) Partido de la Revolución Democrática (PRD)
4) Partido Verde
5) Partido del Trabajo
6) Movimiento Ciudadano
7) Movimiento Regeneración Nacional (MORENA)
8) Partido Encuentro Solidario
9) Redes Sociales Progresistas
10) Fuerza por México
96) Ninguno / Anuló voto
97) No votó
98) No recuerda
99) No quiere contestar / El voto es secreto
SOCIODEMOGRAFICOS
P33. Sexo [NO PREGUNTAR]
1) Hombre
2) Mujer
P34. ¿Cuál es tu estado civil?
1) Soltero
2) Casado
3) Divorciado
4) Viudo
5) Cónyuge
99) NC
P35. ¿Cuántos hijos tienes?
1) No tengo hijos
2) 1 hijo
3) 2 hijos
4) 3 hijos
5) 4 o más hijos
8) NS
99) NC
P36. ¿Cuántas personas viven en tu hogar, incluyendo niños pequeños y adultos mayores?
__________
99) NC
P37. ¿Con quién vives? [RESPUESTA MÚLTIPLE, MARCAR TODAS LAS QUE MENCIONEN]
[ENCUESTADOR MARCAR TODAS LAS QUE APLICAN]
Madre ○
Padre ○
Abuelo(s) ○
Tío(s) ○
Amigo(s) ○
Pareja(s) ○
Hermano(s) ○
Hijos ○
Solo ○
Otros ○
85
La entrevista ha terminado. Muchas gracias nuevamente por contestar las preguntas. 
86
Section C. Focus Group Guide
INTRODUCCIÓN
1. Presentación de los facilitadores y de los objetivos del proyecto.
2. Explicación sobre cómo la identidad de los participantes y toda la información que surja 
como parte de la sesión será tratada con absoluta confidencialidad.
a. Se solicitará a los participantes elegir un sobrenombre, y sólo los sobrenombres 
podrán ser utilizados durante la sesión.
b. Se solicitará a los participantes NO MENCIONAR ningún nombre personal de 
personas durante la sesión. Pueden referirse a las personas como “amigos”, 
“hermanos”, etc.
c. Se le pedirá a los participantes que no hablen sobre la discusión durante la 
sesión con otras personas que no formaron parte.
3. Explicación de la dinámica de la sesión:
 i. Solo una persona habla a la vez.
 ii. Si alguien tiene una opinión diferente al resto del grupo, es importante 
que lo diga.
 iii. No hay respuestas correctas e incorrectas, todos los comentarios 
son importantes
4. Dinámica rompehielos y presentaciones: los participantes compartirán el nombre que 
seleccionaron con el grupo, y deberán compartir por qué eligieron ese nombre. Cada 
participante también compartirá sus expectativas sobre la sesión.
5. Leer el consentimiento en fuerte y preguntar a los participantes si se puede audio-grabar 
la sesión.
GUIA DE TEMAS PARA LA SESIÓN
I. USO DEL TIEMPO Y ASPIRACIONES
1. ¿Qué hacen los jóvenes de su edad que viven en su barrio/colonia para divertirse? 
- Música
- Deportes
- Baile
- Fiestas
- Alcohol/drogas
2. ¿En qué trabajan los jóvenes de su barrio/colonia? 
- Les gusta
- Ingresos
- Tienen varios trabajos
- Trabajo ideal
3. ¿A qué aspiran los jóvenes de su barrio/colonia?
87
- Estudiar
- Familia
- Negocio propio
- Vivir en otro lugar
- Parecerse a alguien
- Ser famosos
 --------------------SEPARAR GRUPOS POR SEXO-------------------
II. NORMALIZACIÓN DE LA VIOLENCIA
4. Ahora, les voy a pedir que me ayuden a clasificar estas acciones. El rojo es para aquellas 
que son muy violentas, el amarillo es para las que son poco violentas y el verde es para 
las que no son violentas. (En la clasificación de cada una, que expliquen por qué).
Saludar en la calle a alguien desconocido
Decirle piropos a alguien
Tener sexo con alguien sin su consentimiento
Vender drogas
Criticar a alguien a sus espaldas
Besar a alguien sin pedirle permiso
Golpear a alguien
Saludar en la calle a alguien desconocido
Burlarse de alguien
Ayudarle a un desconocido
5. ¿En qué situaciones puede ser válido usar violencia? 
- Defensa personal
- Darse a respetar
- Cobrar deudas
- Conseguir algo
- Ejemplos en que ellos hayan tenido que usar violencia
6. Pensando en sus amigas y conocidas, ¿qué formas de violencia sufren más las mujeres? 
¿Qué formas de violencia sufren más los hombres? 
- Ejemplos
- A qué lo atribuyen/quien tiene la culpa
7. Pensando en sus amigas y amigos, ¿cómo es la violencia en las escuelas? 
- Ejemplos
- A qué lo atribuyen/quien tiene la culpa
8. ¿Y cómo es en las familias?
88
- Ejemplos
- A qué lo atribuyen/quien tiene la culpa
----------BREAK DE 10 MINUTOS----------
-----------------------------JUNTAR AMBOS GRUPOS-----------------------------
(ESTIRAMIENTOS)
III. JUSTIFICACIÓN DE LA VIOLENCIA
9. Les voy a contar la historia de Rodrigo. Él tiene 17 años, es de muy bajos recursos y se 
salió de la prepa para trabajar y ayudarle a su mamá con el gasto. Como no conseguía 
trabajo, un vecino le propuso robar pertenencias de otras personas (bicicletas, bolsas, 
celulares) para revenderlas. ¿Qué les parece la propuesta del vecino?
10. Rodrigo decidió entrarle al negocio con el vecino porque les urgía el dinero a él y su 
mamá para pagar la renta. De todas estas personas, ¿a quiénes les parece que sería peor 
que asaltara?
Un empresario
Un obrero
Un político
Una trabajadora de edad avanzada
 
11. Rodrigo se volvió muy hábil asaltando, pudo cubrir los gastos de él y su mamá, y se dio a 
conocer, por lo que el líder de una banda criminal lo invitó a participar con ellos y le 
ofreció protección. ¿Le convendrá unirse?
- Razones por las que sí
- Razones por las que no
12. Rodrigo aceptó unirse y su primera tarea fue secuestrar a un empresario. Al cabo de un 
tiempo, como no estaban pagando el secuestro, el líder de la banda le pidió a Rodrigo que 
asesinara al empresario. Rodrigo lo hizo y la policía lo agarró. Del 1 al 7, ¿qué tan culpable 
creen que es Rodrigo de este crimen?
13. ¿Quiénes más son responsables de este crimen?
El empresario
El líder de la pandilla
La mamá de Rodrigo
89
La policía
La sociedad
El gobierno
14. ¿Su percepción sobre la culpa de Rodrigo en esta historia cambiaría si su familia tuviera 
dinero? ¿Por qué?
15. ¿Quién tiene la culpa de que haya jóvenes que se involucren con el crimen?
- Los jóvenes
- Familia
- Gobierno
- Sociedad (indaguen qué entienden por sociedad)
- Organizaciones criminales
16. ¿Qué propondrían para evitar que más jóvenes se involucren con el crimen o que se 
sumen a bandas?
IV. PERCEPCIÓN SOBRE LA ORGANIZACIÓN
17. ¿Por qué decidieron participar con (nombre de la organización)? 
18. ¿En qué les ha ayudado (nombre de la organización) en su vida personal? ¿Qué han 
aprendido? 
- relaciones familiares
- capacitación para el trabajo
- manejo de emociones
- uso de sustancias
- poner en práctica
19. ¿Qué le cambiarían al programa? ¿Cómo lo mejorarían?
20. ¿Creen que este programa puede servir para prevenir la violencia en jóvenes?
90
Section D. Correlations Between Survey Violence Strata and
Sociodemographic and Normalization Variables 
The figures in this section show correlations between important survey variables and our violence 
strata. We start by analyzing correlations between our individual-level variables (as measured by 
the survey), and then we present correlations between our strata variables and individual-level 
victimization, as well as with our normalization variables. Overall, correlations seem to suggest 
that normalization of violence is more likely to be influenced by specific individual-level 
experiences and not so much by recent community-level violence. Results in this section should 
be interpreted with caution. Simple correlations do not provide us with causal conclusions and 
do not control for possible other confounding factors or omitted variables. Relationships in this 
section should only be interpreted as associations without considering controls. Figure 1 
presents the percent victimization, separated by gender and perceived community security, 
showing that individual victimization is correlated with perceived security level.
Figure 1. Percent of victimization, by gender and perceived community security
The following figures show whether our individual-level variables are correlated with municipal 
strata. Figure 2 presents correlations between municipal-level strata elements (homicide rate, 
perceived security, and victimization) and individual-level victimization across a variety of crimes, 
showing little correlation across elements. However, this figure demonstrates some level of 
correlation between victimization across crimes. Being hurt by a weapon (“shot”) is correlated 
with attempted homicide at the 0.36 level. Further, experiencing robbery and car robbery are 
correlated at the 0.21 level. 
Figure 2. Correlations between municipal-level strata elements and individual-level 
victimization across different crimes
91
Figure 3 demonstrates correlation between municipal-level strata components and individual￾level normalization of violence across certain acts. Strata levels are not highly correlated with 
normalization of violence across elements. However, there are strong correlations between 
various normalization-of-violence elements. Notably, responses regarding nonconsensual sex and 
hitting an individual are correlated at the 0.47 level, hitting on an individual and nonconsensual
kissing at the 0.52 level, and criticizing looks and making fun of an individual at the 0.68 level. 
Figure 3. Correlations between municipal-level strata components and individual-level 
normalization of violence across different acts
92
It is important to keep in mind that these correlations are not causal findings and should not be 
interpreted as such. The correlations show that areas with high levels of violence do not 
necessarily correlate with higher victimization or normalization of violence. However, the analysis 
does show correlations between victimization and perceived security and, most importantly, 
between victimization across crimes and between various normalization-of-violence elements. 
Together with the findings above, the correlations seem to suggest that normalization of violence 
is more likely to be influenced by specific individual-level experiences and not so much by recent 
community-level violence. This is beyond the scope of our analysis, but more fine-grained data 
on individual exposure to different forms of violence over time would improve our understanding 
of the mechanisms that generate a normalization of acts of violence. 
93
Section E. Additional Results on Program Participation and Violence 
Perceptions
In Table 1, we observe that participation in prevention programs is negatively correlated with 
risk perceptions about gang belonging. As expected, past participation in violence prevention 
programs is associated with a decrease in the perception that gang membership is fun. 
Participation in drug prevention programs is associated with an increase in the perception that 
gangs increase one’s security and are fun, and past participation in arts and culture groups is 
associated with a decrease in the perception that gang membership will lead to imprisonment. 
Notably, this is not a causal relationship; it is possible that those who participate in drug 
prevention programs are more likely to hold these views in the first place.
Table 1. Gang Risk Factors and Past Program Participation
94
Section F. Additional Experimental Results
Table 1 examines the results for individual blame attribution parsing apart the sample by 
perceived insecurity level—high vs. low. If the perpetrator is from the middle class, he is more 
to blame in both high- and low-perceived-insecurity areas. In low-perceived-insecurity areas, as 
the crime increases in severity, the individual is more to blame, and while this estimator has the 
opposite sign in high-perceived-insecurity areas, it is not significant. In both types of community, 
the perpetrator is more to blame if he is a gang leader.
Table 1. Blame for Individual by Perceived Security Level
Table 2 examines the results for individual blame attribution parsing apart the sample by 
victimization level—high vs. low. If the perpetrator is from the middle class, he is more to blame 
in both high- and low-victimization areas. Again, In low-victimization areas, as the crime increases 
in severity, the individual is more to blame, but the effect is not significant for high-victimization 
areas. Notably, in high-victimization areas, if the victim is a worker vs. a businessman, the 
individual is more to blame. In both types of community, if the perpetrator is a leader of a gang 
vs. a follower, he is more to blame.
Table 2. Blame for the Individual by Victimization Level
95
Table 3 presents the results for individual blame attribution breaking the sample by respondent 
SES, and Table 4 presents the results for individual blame attribution breaking the sample by 
respondent gender. Both SES groups blame the individual more if he is of middle vs. lower class.
Both men and women blame the individual more if he is of the middle class as compared to the
lower class. Crime severity does not matter in any of these samples, and in every case, if the 
perpetrator is a gang leader he is more to blame than if he were following orders. For men, if the 
victim is a local politician, the perpetrator is more to blame. 
Table 3. Blame for Individual by Respondent SES
Table 4. Blame for Individual by Respondent Gender
96
Table 5 presents the results for blame attribution to the individual vs. the government and 
society parsing out the sample by perceived security level; it shows very similar results to the 
model parsing apart the sample by homicide level. 
Table 5. Blame for Individual vs. Government + Society by Perceived Security Level
Table 6 presents the results for blame attribution to the individual vs. the government and 
society dividing the sample by victimization level of the areas from where the respondents come, 
with slightly different results from other examinations of community security contexts. It is those 
in low-victimization areas who attribute more blame to the individuals if they come from the 
middle class. Crime severity has a significant and divergent effect—those in high-victimization 
communities attribute more blame to the government and society for more severe crimes, while 
those in low-victimization communities attribute more blame to the individual for more severe 
crimes. Further, if the victim is a worker vs. a businessman, in high-victimization areas, more 
97
blame is attributed to the individual. In both cases, if the perpetrator is a gang leader, more blame 
is attributed to the individual. 
Table 6. Blame for Individual vs. Government + Society by Victimization Level
Table 7 examines differences by gender. Both genders attribute more blame to the individual if 
the perpetrator is of the middle class as compared to the lower class. Men attribute more blame 
to government and society if the victim is a local politician; women do not. Both attribute more 
blame to the individual if the perpetrator is a gang leader. 
Table 7. Blame for Individual vs. Government + Society by Gender
Table 8 examines the differences by respondent SES. Those of high SES attribute more blame 
to the individual if he is of the middle vs. the lower class; no results exist for those of low SES. 
98
Crime severity and victim identity are not significant factors affecting blame for either group. For 
both groups, if the perpetrator is a gang leader, more blame is attributed to the individual.
Table 8. Blame for Individual vs. Government + Society by Respondent SES
99
Section G. Percentage of Youth Who Have Been Victims of Domestic Verbal 
and Physical Abuse 
The table below suggests that reports of abuse are more common among lower-SES respondents.