DYNAMICS OF VIOLENCE IN

by

Viveca Pav´onTercero

APPROVED BY SUPERVISORY COMMITTEE:

Jennifer S. Holmes, Chair

Clint Peinhardt

Patrick T. Brandt

Dohyeong Kim Copyright c 2018

Viveca Pav´onTercero

All rights reserved This dissertation is dedicated to the Two that convinced me I could. DYNAMICS OF VIOLENCE IN EL SALVADOR

by

VIVECA PAVON´ TERCERO, BS, MBA

DISSERTATION

Presented to the Faculty of

The University of Texas at Dallas

in Partial Fulfillment

of the Requirements

for the Degree of

DOCTOR OF PHILOSOPHY IN

PUBLIC POLICY AND POLITICAL ECONOMY

THE UNIVERSITY OF TEXAS AT DALLAS

August 2018 ACKNOWLEDGMENTS

This research would not have been possible without those organizations that believed in my work and contributed financially for its completion. A special thanks to CUSLAI for their support during my studies and their help with my research. Without their travel grant this dissertation would have been quite difficult to complete. Dr. Monica Rankin and Dr. Jennifer Holmes were outstanding during our field work, I learned a lot from their interviewing skills. I would also like to thank Phi Kappa Phi Cyrus D. Cantrell III Academic Travel Grant for helping fund my research in El Salvador.

To my committee, thank you for your insight, I have learned so much from each of you. Dr. Holmes, thank you for continuously working with me and helping me expand my view on the ‘dark side’. I would also like to thank Dr. Brandt for his direction in conducting research; Your training has been invaluable. Dr. Peinhardt, I truly appreciate your guidance and how you have encouraged me to think beyond the theoretical norms. Lastly, Dr. Kim, thank you for helping me link crime policy to GIS.

I am grateful to my family and friends for the support and encouragement throughout the years. Thank you Asli, for not dropping the subject until I did something about it. Now I know how difficult it truly is to earn that two letter introduction. I would like to thank Alisha for being unapologetically you. Meeting you that first day of class is something I will forever be thankful for.

Le quiero agradecer a mi mami por su apoyo incondicional aun cuando no sabia que lo necesitaba. Gracias por recordarme que hay que luchar por las cosas buenas de la vida.

I would like to thank dad and my brothers, for the endless conversations on how we could fix things. Thanks for listening and never thinking (out loud at least) that I was too young or naive to participate. Thank you guys for being a constant source of inspiration. A special thanks to my blue blood moon, for being everything.

v Jeremy, I am grateful for you every day. Your love and support has meant the world, this is our accomplishment. Know that I am forever changed because of who you are. You and the pups make it all worth it.

June 2018

vi DYNAMICS OF VIOLENCE IN EL SALVADOR

Viveca Pav´onTercero, PhD The University of Texas at Dallas, 2018

Supervising Professor: Jennifer S. Holmes, Chair

Central America is the deadliest region in the world. The UNODC reports that the level of violence in this region is higher than in any single nation, including those at war. El Salvador’s 2015 homicide rate was listed as the highest for any country in nearly 20 years. This research seeks to better explain policies within El Salvador and how they continue to affect gang activities and crime. The research also seeks to fill in the gap on gang network dynamics and their level of power in affecting policies.

I begin by exploring the policies implemented in the early 2000s and explaining how they have had a significant effect in gang development but also in political gain for acting parties. In an attempt to maintain control over a territory state leaders may chose policies that favor a public appearance of control. This study seeks to address if a country is willing to implement ’punitive populism’ in exchange for votes by analyzing the case of El Salvador. I argue that despite numerous failed attempts at controlling violence through punitive policies the country continues to enforce these actions and use them as platforms for future elections.

Chapter three in this research uses social network analysis (SNA) to identify the temporal relationship between paired municipalities according to homicide. This method looks at the interconnectivity of homicide counts from one year and the next between municipalities, meaning that it is evaluating how two seemingly distinct regions can be responding to each other when it comes to homicide rates. The idea is to identify which municipalities are

vii responding to crime in other municipalities around the country. This becomes a critical aspect of violence in the country because gangs can react to a specific event by attacking regions where rival gangs operate, not necessarily adjacent areas. In this study I conducted a single datum correlation coefficient (SDCC) to create this network connection. Using SNA we can better identify what regions of the country are of interest during a specific point in time and can help policy makers establish areas needing additional backing. This novel method introduces a new way of observing criminal behavior and helps identify hubs of criminal activity as well as vulnerable relationships among municipalities which could be indicative of gang retaliation areas.

The 2012 Peace agreement in El Salvador between Salvatrucha (MS-13) and Barrio 18 was an unprecedented truce between two of the deadliest rival gangs in Latin America. This agreement serves as an indication of cooperation with a greater purpose between groups that had been deemed unorganized. This raises concerns of what circumstances will facilitate cooperation between rival groups and what can be expected of this collaboration. Following the truce there was a genuine fear that gangs in El Salvador could become political actors by using Violent Lobbying and/or Violent Corruption. Even though the gangs have not proven to have political power there is still a fear that they can continue to influence politicians behind closed doors. This paper seeks to identify the tactics the gangs could use if they did attempt to lobby the state. These specific violent acts help us understand what pressures the government of El Salvador could have been facing that led to the truce being agreed upon. It also distinguishes what changes in violence the country experienced during and after the agreement. In order to identify these specific patterns I use event data analysis on newspaper stories from El Diario de Hoy to help identify is any coercive methods where used by las maras.

These three chapters evaluate different aspects of violence in El Salvador originating from gang activity. I begin with a historical approach of policy implementation help introduce

viii how the gangs developed and how they gained power to be able to influence violence beyond the reach of direct contiguity. I conclude by presenting past uses of violence with a direct purpose in order to understand the level of political power the gangs could obtain for future gain.

ix TABLE OF CONTENTS

ACKNOWLEDGMENTS ...... v ABSTRACT ...... vii LIST OF FIGURES ...... xi LIST OF TABLES ...... xiii CHAPTER 1 INTRODUCTION: AN OVERVIEW OF GANGS IN EL SALVADOR 1 CHAPTER 2 PUNITIVE GANG POLICIES FOR POLITICAL GAIN ...... 19 CHAPTER 3 THE IMPACT OF GANG NETWORKS ON HOMICIDES IN EL SAL- VADOR ...... 44 CHAPTER 4 ESMED - EVENT DATA OF MARAS IN EL SALVADOR ...... 66 CHAPTER 5 CONCLUSION ...... 92 APPENDIX A BIC & RSS ...... 97 APPENDIX B NEWSPAPER QUERY ...... 98 APPENDIX C EVENT DATA ON CONVERTED HTML FILES ...... 99 REFERENCES ...... 102 BIOGRAPHICAL SKETCH ...... 110 CURRICULUM VITAE

x LIST OF FIGURES

2.1 Prison Population ...... 28 2.2 Incarcerated Gang Population ...... 29 2.3 Monthly Homicide 2002-2017 with significant structural breaks ...... 32 3.1 Homicide in Mexico and ...... 48 3.2 Homicide in El Salvador 2005-2013 ...... 49 3.3 2013 Homicide Rate by Department ...... 50 3.4 2013 Homicide Rate by Municipality ...... 51 3.5 SDCC Network Graphs ...... 56 3.6 Geo-referenced Homicides 2006 ...... 58 3.7 Global Moran’s I 2006 ...... 59 3.8 Hot Spot Analysis 2006 ...... 60 3.9 Cluster and Outlier Analysis 2006 ...... 61 3.10 SDCC for 2007 ...... 62 4.1 Homicide and Homicide by Gangs ...... 70 4.2 Police Officers Murdered ...... 71 4.3 Eventus ID Coding Process ...... 74 4.4 El Diario de Hoy newspaper sheet sample ...... 76 4.5 web2eventus Output ...... 77 4.6 Example of Actor Coding ...... 79 4.7 Example of Verb Coding ...... 81 4.8 Example of Location Dictionary ...... 82 4.9 Example of Raw Output ...... 83 4.10 STATA Geocoded Output ...... 83 4.11 Main Source of the Events ...... 85 4.12 Main Action of the Events ...... 86 4.13 Main Target of the Events ...... 86 4.14 Main Location of the Events - State Level ...... 87 4.15 Validation - GSR v EVENTUS ID ...... 88

xi 4.16 ESMED v OCVED News Sources ...... 90 4.17 Homicide Distribution ...... 91 A.1 BIC & RSS (2002-2017) ...... 97 B.1 Newspaper Queries ...... 98 C.1 Main Source of the Events ...... 99 C.2 Main Action of the Events ...... 100 C.3 Main Target of the Events ...... 100 C.4 Main Location of the Events - State Level ...... 101

xii LIST OF TABLES

1.1 Elements associated with Central American gangs ...... 7 1.2 Gang Role Classification ...... 11 1.3 Features of the Maras in El Salvador ...... 12 2.1 ...... 33 2.2 Mano Dura policies ...... 42 3.1 Data Comparison ...... 54 3.2 Sum of Correlated Municipalities ...... 57 4.1 Actor Dictionary Categories ...... 78 4.2 Verb Categories ...... 80

xiii CHAPTER 1

INTRODUCTION: AN OVERVIEW OF GANGS IN EL SALVADOR

Central America is the deadliest region in the world. The UNODC reports that the level of violence in this region is higher than in any single nation, including those at war. In 2017, Central America had an annual homicide count of 14,575 people, with 13,129 of these casualties in El Salvador, , and − the Northern Triangle − (EFE, 2018). This staggering number is far beyond even the average civil war death toll of 10,500 (Lacina, 2006). The countries of the Northern Triangle have experienced war in recent decades and are unfortunately familiar with high homicides rates, yet current homicide figures are alarming. El Salvador’s 2015 homicide rate of 108.6 people per 100,000, for example, is the highest for any country in nearly 20 years (Bank, 2018). Gangs in El Salvador El Salvador is a country with a long history of violence resulting from a civil war (1980- 1992) that killed 70,000 people and displaced over a million (Baylora, 1983; Buergenthal, 2004; Wood, 2003; Perez, 2003; (ICG), 2017). Even though there was an end to the war violence levels have not decreased. Drug Trafficking Organizations (DTOs) and gangs are responsible for most of the bloodshed in the Northern Triangle (Demombynes, 2011). Al- though there is some trafficking through El Salvador the country can attribute the majority of its crime and violence to gangs (maras in Spanish) (Garzn, 2014). This is not surprising given that it has become an epicenter of gang activity (Cruz et al., 2000; De Cesare, 1998; Dudley, 2014; Smutt and Miranda, 1998). Today, gangs have around 60,000 active members in El Salvador alone ((ICG), 2017). There are two prominent gangs in El Salvador today: Mara Salvatrucha (MS-13) and the (Barrio 18), which is currently divided into two factions: The Revolu- cionarios (Revolutionaries) and the Sureos (Southerners) (Cruz et al., 2017). Their origins, though, are far from the small Central American country. In the 1970s and 1980s many

1 residents of El Salvador emigrated due to growing concerns of armed conflict and the up-

surge of armed opposition. They fled towards Sweden, Australia, Canada, and ,

but the majority sought refuge in the United States, particularly in (Mar-

tinez D’aubuisson, 2015). Economic hardships, social exclusion, violence, and the beginning

of what would eventually become a civil war in El Salvador were some of the pressing issues

that led so many to leave the country (Gmez Hecht, 2017). Unfortunately, many Salvadoran

immigrants encountered similar issues in the United States.

Once in Los Angeles many of the young males found it difficult to assimilate to American

culture. It was evident that they were different from the other Spanish speaking groups in

the city. They faced discrimination and marginalization leading them to embrace their Sal-

vadoran heritage and tighten their bond. Anthropologist Tom Ward dates the first meetings

of what would eventually become the Mara Salvatrucha to 1979 (Wards, 2012). They did

not resemble other gangs in Los Angeles so instead they started to meet and call themselves

a mara. The word mara is short for marabunta: a plague of ants or a crowd. Marabunta

was used as a part of a tag-line for the movie The Naked Jungle in 1954 which, after being translated to Spanish in the sixties became a box office hit in El Salvador. The term was shortened and adopted colloquially as mara to refer to a group of friends that, like ants, will protect each other (Martinez D’aubuisson, 2015; Grillo, 2016). The use of the word mara at

this point was quite different from what it means now. It was also quite different from the

definition of a gang. Klein and Maxson define a street gang as “any durable, street-oriented

youth group whose involvement in illegal activity is part of its group identity” (Klein and

Maxson, 2010). The word salvatrucha is a combination of two local words: ‘salva’ is short for Salvadoran while ‘trucha’ is slang for someone smart, cunning, or alert (Justice, 2017).

The first Salvatrucha members were young immigrants wearing tight pants, black t-shirts with long hair and a rocker look; they were known as the ‘Mara Salvatrucha Stoners’ (MSS)

(Martinez D’aubuisson, 2015). Mara Salvatrucha gained popularity with new immigrants

2 coming from El Salvador. Their distinct look and overall culture made them a distinguish-

able group in Los Angeles, making them easy police targets. By 1983 they had established

themselves as a group with standardized entry rituals and rivals.

Meanwhile, in El Salvador, the murder of Archbishop Oscar Arnulfo Romero in March

of 1980 marked the beginning of the civil war. The war was a struggle between farmers and

the elites who were backed by the current government. When the farmers engaged in a 1979

strike demanding better working conditions and higher salaries, the government responded

by allowing the National Guard to use military force. The emerging violence enraged the

farmers and led them to form left-wing guerrilla forces (Wood, 2003). By October of 1980,

five leftists guerilla groups merged to form the Farabundo Mart National Liberation Front

(in Spanish: Frente Farabundo Mart para la Liberacin Nacional, FMLN ) (Martinez et al.,

2013). FMLN had an arduous recruitment campaign, creating youth organizations made up of high school students that were later trained for military offenses. As the conflict gained momentum more individuals emigrated fearing for their lives.

Back in Los Angeles, Mara Salvatucha Stoners openly welcomed young men that had served as guerillas or child soldiers (Grillo, 2016). These men had experienced horrific acts of war. They embraced gang rivalries and many were quickly arrested. Within the prison system MSS members were introduced to La Mafia Mexicana also known as ‘la Eme’ (the M).

This gang was made up of members of other gangs that had been recruited in prison where they were expected to fight and gain some form of social power. This gang earned respect and loyalty by protecting all Latino gang members within prisons. Gang affiliations inside the prison system were about ethnic grouping instead of nationality. The gang capitalized on this notion and gained influence over most Chicano gangs in Los Angeles. La Eme identified itself with the number 13, given that the letter M is the thirteenth letter of the alphabet

(Martinez D’aubuisson, 2015). Gangs outside of the prison system adopted the 13 as a part of their name to indicate submission to la Eme. Their reach became legendary as the

3 gang that controlled all other gangs from within prison walls. Over the next few years MSS members started to assimilate into Chicano life and as they experienced prison in L.A. they adopted the Eme last name. By 1986 the Salvadorian gang in Los Angeles was known as

Mara Salvatrucha 13 (MS-13). With a heavy influx of fleeing conflict back home,

MS-13’s membership grew quickly. Their behavior evolved from fist fights with rival gangs to knife and machete fights (Dudley and Silva Avalos, 2018).

The 18th Street gang was already a well established street gang in Los Angeles. It originated as a faction of one of the most veteran gangs in L.A., the Clanton 14 in the late

1940s (Martinez et al., 2013). The gang was well known and had clicas, or chapters, in many cities in the United States and Latin America where it was better known as ‘Barrio 18’ or simply 0M-180. In El Salvador a clicas called Los Malditos welcomed Carlos Ernesto Mojica

Lechuga as a member in the late 1970s and gave him the street name Lin, now know as el viejo Lin the public speaker for Barrio 18 (Martinez et al., 2013). Back in L.A. Barrio

18 and MS-13 were allies. Barrio 18 had also experienced discrimination by the traditional

Chicano gangs in the city because they had Filipino, Caribbean, and Central American members. When the newly formed MS-13 needed a partnership to help gain street credit and momentum, Barrio 18 was eager to help.

The reason for the fallout between the two gangs is a bit of a mystery. There are accounts of arguments over territory, fights between leaders of both sides, and even accounts of arguments over women. What both gangs do agree on is that their hatred for each other materialized in 1989 when a group from Barrio 18 killed Shaggy, a member of MS-13, in an alley parallel to the Martin Luther King Boulevard in L.A. (Martinez et al., 2013). Shaggy’s death erupted a full war between these gangs that transcended geography and time.

By the beginning of the 1990s both gangs had a violent reputation and were at odds with the police. Following the Rodney King riots California passed tough, anti-gang laws which led to many incarcerations. Federal laws regarding immigration had favored citizens

4 of El Salvador by granting them temporary protected status due to armed conflict. This favor was withdrawn for those individuals that did not uphold the law and changes were being made to the law that made it easier to deport felons after the completion of their sentence. The Anti−Drug Abuse Act of 1988 introduced the concept of aggravated felony, these crimes were declared deportable under the immigration statute. The act listed crimes such as murder, drug trafficking, and trafficking of other illegal goods as aggravated felonies

(TRAC, 2006). The 1990 Immigration Act added money laundering and crimes of violence to the aggravated felony category. Incarceration of gang members resulted in deportations.

After the civil war ended 1992 El Salvador was able to start experimenting with its own policy changes. The signing of the Chapultepec Peace Accords which enabled FLMN to become a political party. The agreement also called for a reduction of the military and a transfer of power from the police to a newly formed civilian police force ((ICG), 2017;

Matrix, 2015). After twelve years of war the country was in ruins. There was crumbling infrastructure, weak government institutions, and a broken social fabric due to unresolved disputes. The war brought with it unemployment; most of the men that took arms were now unemployed former guerrilla or military. Many of these men had been wounded and found it even harder to find a source of income. There was an increase in robberies and kidnapping.

Neighborhood gangs proliferated across the country (Martinez D’aubuisson, 2015). That same year the INS Violent Gang Task Force was created in the United States. Working with local police agencies the federal government launched a deportation campaign against gang members (Justice, 2018). The Illegal Immigrant Reform and Immigrant Responsibility

Act (IIRIRA) of 1996 accelerated this by requiring the deportation of undocumented gang members after their sentence had been served (TRAC, 2006). Hundreds of Californian gang members were deported.

Boys and young men that had left El Salvador were returning years later as men. They all shared gang life, either in MS-13 or Barrio 18, and having spent time in prison in California.

5 These were the men that transformed local neighborhood gangs into transnational criminal

groups, las maras. They had little in common with the locals and found refuge in gang life.

Mareros, a gang member in Spanish, took over neighborhoods and demanded allegiance to their respective gang from smaller groups. New clicas emerged with creole names. Mara

Salvatrucha 13 and Barrio 18 prospered in El Salvador, and by the end of the decade every

poor neighborhood and shanty town had been claimed (Martinez D’aubuisson, 2015).

Factors influencing Mara growth

Criminal deportees returned to Central America and encountered a hostile environment.

The country they left as children and young men did not have a welcoming social environ-

ment, especially not for them. They returned as criminals and were not desired by Salvadoran

society. Area specialist Jose Miguel Cruz defines gangs in Central America as the outcome

of a permanent social dynamic where gang members, society, and the state are in a contin-

uous relationship that dictates criminal behavior. It is the overlap of the three groups that

determines if and how a gang will develop (Cruz, 2005). El Salvador was experiencing social

and political changes in the 1990s that proved to be detrimental for a stable relationship be-

tween its groups. Cruz identified ten elements associated with gang proliferation in Central

America, divided into four relational categories: see Table 1.1.

Each of these categories details a set of conditions that directly affect youth behavior

and their involvement in gangs. It is worth noting that poverty is not listed. Even though

gangs do flourish in poorer neighborhoods, poverty itself is not considered a causal factor for

joining a gang. Instead, elements such as precarious socio-economic conditions, few education

opportunities, unemployment, and a community lacking basic services such as electricity or

running water can lead to feelings of social exclusion which can cause a young individual to

seek gang affiliations (Cruz, 2004, 2005). This was, and continues to be, a relevant factor

in El Salvador. Unfortunately, deportations from the United States were happening at the

same time that the country was trying to rebuild itself following the civil war. This high level

6 Table 1.1. Elements associated with Central American gangs Relational level Causal category Social Social exclusion Culture of violence Migration Rapid and disorganized urban growth Community Drugs Lack of social capital Family and friends Domestic problems Friends belonging to a gang Dynamics of violence Individual Lack of individual or collective identity Source: adapted from (Cruz, 2005) of migration, in the form of deportations, is a well documented causal factor for the presence of gangs in Central America (Ribando, 2005; ZIlberg, 2004). El Salvador had smaller gangs fighting over territories before deportations began. These disputes were limited to street level arguments and were not a top priority for citizen security. Once deportees arrived with distinct gang features like their different Spanish dialect, and distinctive tattoos the externalization of gang affiliation became evident. Smaller gangs adopted this behavior and discovered the importance of gang identity (Cruz, 2005). Migration is only one of the elements that helps explain gangs in the area. Another condition associated with gangs is rapid and disorganized urban growth. A lack of urban organization causes cities to be built without proper access to basic services and without public spaces. Increased migration is associated with this rapid growth that in many cases increases the size of a household beyond is capacity. A study on ’s metropolitan area found that over-crowded homes result in more children and youth in the streets (Smutt and Miranda, 1998). The last social relational level affecting gang presence in Central America is a culture of violence, defined as a system were values, norms, and attitudes encourage and even legitimize the use of violence in intra-personal relationships (Huezo, 2001; Martin-Baro, 1992). Homicide is the clearest indication of a culture of violence: Central America has experienced higher rates than most

7 other Latin American countries since the seventies. Police reports in El Salvador estimate that 65% of the murders in the country are due to “social violence” (PNC, 2003) which is violence motivated by benefit or social power and can be expressed at multiple levels of society. This type of violence manifests itself as domestic violence, street fights, interpersonal violence and is evident across all age groups (Moser and Winton, 2002).

Community relational levels such as a lack of social capital are also causal factors for gang involvement. In communities where there is little trust among neighbors there is less socializations. Disputes tend to be dealt with more violently than in circumstances where there is some form of collective action in favor of the greater good. Communities that had recently faced civil conflict are also burdened by competition for survival (Cruz, 2004).

This was the case of many communities in El Salvador where neighbors found themselves on opposing sides of the war. Unfortunately when deportees tried to establish community relationships their presence only exacerbated the issues of distrust. Another element serving as an explanatory factor for gang presence and even expansion is the availability of drugs and illicit markets. Drug trafficking is commonplace in the Northern Triangle, with about

90% of the cocaine entering the United States (U.S.) from South America passing through the Central American-Mexico Corridors, and 80% of this amount first entering a Central

American country before being smuggled into Mexico (INCSR, 2014). The consumption of drugs is not a necessary condition for belonging to a gang but in many cases the illicit markets for it are controlled by gangs. This is a growing problem in the region but one that has been in place for decades.

Other factors that serve as explanatory elements of gang presence are more personal. Be- ing a member of a dysfunctional family, such as single-mother households and grandparents raising children, has a big impact on potential gang affiliation (Cruz, 2005). Abandonment, neglect, and abuse from within a home are important conditions that lead a young man into gang life. Gangs provide a sense of belonging and protection, they seek members that will

8 devote themselves to the group and gain respect and affection from their peers. This is in many cases what fills the void left by dysfunctional families. Having a family member of a friend belonging to a gang is another important component of gang presence because it is associated with social pressures to join. These two elements were evident in El Salvador when gangs were gaining momentum. Many families had been torn apart by the war and migration. Society was vulnerable and many families were experiencing different levels of dysfunction. As gang presence grew so did the potential of knowing someone that belonged to one and so gang proliferation continued. The saying “violence begets violence” is exem- plified in gangs. Violence is a part of gang identity and a requirement for belonging to one.

When a young man wants to show his devotion to the gang he has to pass a violent test.

In many cases this is an act against civil society or members of opposing gangs that will retaliate with equal or greater force. Violence becomes a cycle hard to break. This is the case of MS-13 and Barrio 18. This is another element that affects gangs in El Salvador. On a personal level a lack of identity is significant indicator of vulnerability to gangs. A lack of role models limits the available options for young men to look up to. For many, gang life is the only possible alternative for “success”. Gangs offer the possibility of financial, material, and even emotional stability that young men desire. Which is for many young individuals the only visible option for support.

The ten elements previously described on their own have a negative effect on society but together can have an affect the social fabric of an entire country. El Salvador was a country weakened by its internal disputes with the issue further exacerbated by the influx of criminal deportees from the United States in the 1990s. Social, community, and personal issues all facilitated the proliferation and continuous growth of Mara Salvatrucha 13 and Barrio 18.

But the gang phenomenon is dynamic and it is important to highlight that the Maras of the early 1990s are the not the same as what we experience today (Cruz, 2005). As the elements affecting gang presence change gangs have adapted to continue developing.

9 Gang Features A typical marero in El Salvador is a young male about 25 years old from a low-income family in a marginalized neighborhood without a secondary education. Surveys conducted in El Salvador over the last few decades have documented the evolution of gang characteristics both on the streets and in the prison system. The first set of surveys were conducted with youth in the streets in 1996 and 2001 (Cruz and Portillo Pea, 1998; Santacruz and Concha-Eastman, 2001), while the subsequent surveys were conducted in prisons (Cruz and Carranza, 2006; Aguilar and Carranza, 2008). Despite having different methodologies their results have similar conclusions. The average age for joining a gang is fifteen, the vast majority of members do not have a secondary education, and over half of them dropped out of school before joining the gang. Most gang members are unemployed and over 80% of them have never held formal employment (Cruz et al., 2017). Even though more than half of the members are under the age of 25, gangs today have a larger share of older members than ever before. Economic insecurity is a common characteristic for most gang members with the majority of them belonging to low-income families with a monthly income under $250 (Cruz et al., 2017). Gang Organization The structure within gangs is constantly fluctuating due to the nature of the gangs themselves. Death and incarceration make them a dynamic organization. Every position has an important responsibility and they operate in a system that allows for climbing the organizational ladder. It is estimated that the majority of gang members (76.2%) hold regular positions while less than 10% of members have leadership roles (Cruz et al., 2017). Table 1.2 is a classification of gang positions and details about some of these roles. Both maras in El Salvador have a similar classification for their members, only a few roles have different meaning. These have been outlined below. The hierarchical organization differs between the gangs. MS-13 is the more organized of the two. It has different levels of management, they divide their groups by geographic

10 Table 1.2. Gang Role Classification Position Role Details Leadership Position Palabrero clica leader Ranflero National gang leader Corredor Clique Leader Encargado Segunda Palabra Segundo Corredor Regular Members Marero gang member - only for MS-13 Pandillero gang member - only for Barrio 18 Homeboy Clica or gang companion (male) term of endearment Soldado Activo Brincado Artillero Gatillero Sicario Ejecutor Collaborators/Informants Informante Collaborators are not quite or never will be gang members Poste Paro Chequeo Haina girlfriend/wife of gang member Movedor drug/arms movers Other Calmado No longer active Wannabe Aspiring members Criteriado Protected Witness Source: Author’s creation adapted from (Cruz et al., 2017; Dudley and Silva Avalos, 2018) distribution with the clica at the street or neighborhood level. When clica expand beyond their original neighborhood they are called sectores (sectors), these operate as a franchise of the original clica. Clicas are the grouped into programas (programs) at the regional level and the ranfla at the national level. The ranfla is divided into a group making decisions from within the prison system and a group operating on the street. These two groups manage MS-13’s structure and make the most important decisions (Cruz et al., 2017). Ranfleros are

11 national leaders, they send orders to the Palabreros who are clica leaders. These then send the various soldiers to meet the order. They follow a top-to-bottom chain of command in El

Salvador.

Barrio 18 is not a well organized gang. The gang is divided into two independent factions: the Revolucionarios (Revolutionaries) and the Sureos (Southerners). These groups have little organizational structure; members themselves have a difficult time identifying a leadership pattern. The only clear organization is the geographic division of hierarchy. Neighborhood and city groups are the smallest units in Barrio 18, these are called canchas. Tribus are a collection of canchas at the regional level. The Palabreros are the top gang leaders, they operate from within the prisons. Their contacts outside of prison are the Palabreros “en la libre” (on the street), they are the ones that make sure orders from prison are carried out. Even though there are cancha leaders these are more concerned with local issues and maintaining power. Because Barrio 18 does not have a clear hierarchical structure of power its two factions have had violent disputes among themselves.

Table 1.3. Features of the Maras in El Salvador Feature MS-13 Barrio 18 Origin Los Angeles Los Angeles Members N/A 70,000 clicas 249 28 Incarcerated members Names Mara Salvatrucha 13 Barrio 18 Mara Salvatrucha Pandilla 18 MS-13 M-18 18th Street gang Women Members Unofficially Not Anymore Leadership El Viejo Lin Factions None Revolucionarios (Revolutionaries) Sureos (Southerners) Source: Author’s creation adapted from (Cruz et al., 2017; Dudley and Silva Avalos, 2018; Amaya and Martinez, 2015)

12 Gang Rules In order to maintain some form of power the gangs have established a set of guidelines. These are internal rules specific to each gang but they overlap in many instances:

• Letters and numbers. “Las letras” (the letters) reference the letters M and S, “los numeros” (the numbers) refer to the number 18. Neither gang is allowed to speak about the other and is to respect their affiliation above all. This policy is very important because it is through these letter and numbers that the gangs tag their territories. It is a sign of allegiance and a centering point for the gangs (Savenije, 2009). Hatred for each other is part of what holds each gang together giving it a sense of purpose (Amaya, 2011).

• No theft. There are two reasons for this rule: extortion and community building. Gangs base their economy on extortion and they cannot have internal competition in the form of theft. In many cases the extortion is accompanied by a promise of protection from rival gangs and other criminal groups, included protection from theft. The second reason for the no theft rule is community building. Gangs that wish to build relationships with certain communities have to protect them in any way. It should be noted that no theft means that an individual gang member can only engage in theft when palabrero has authorized it (Dudley and Silva Avalos, 2018).

• No women in the gang. Barrio 18 allowed women gang members until the year 2000 but gang leaders changed their view on women, considering them vulnerable and susceptible to sharing information with the authorities. Women members were removed from Barrio-18 official ranks and no new women are allowed. MS-13 on the other hand realized that the only people able to visit gang members in prison were women. They were then given a role as messengers and eventually evolved into members themselves. Women have also become valuable for cashing out extortions since they do not look

13 like a traditional gang member thus facing less scrutiny and draw less attention. Even

though this rule has technically changed for MS-13, the value placed on women is not

equivalent to that of a man (Cawley, 2013).

• No rape. There is little clarity about this rule especially because there are reported

events of female gang members being gang raped as their initiation (Wards, 2012).

This rule seems to apply to female companion gang member relatives exclusively. An

unspoken origin for this rule may be from within the jail system where rape of younger

incarcerated members is known to happen without punishment (Amaya and Martinez,

2015).

• No Crack. The gangs do not have a moral issue with the use of drugs but they do not

want the use of any substance that without their supervision could lead to a member

becoming an informant. Because crack cocaine is considered the most harmful drug

in this regard both gangs prohibit its use. Some clicas have set rules that only allow

members to smoke marijuana or drink in gang-only events to limit the possibility of

unsupervised sharing (Dudley, 2015). Another reason for opposing the use of this drug

is its addictive nature. Gangs are concerned that they can lose a member’s loyalty

because of them.

• No snitching.

• No leaving. Gang members are not allowed to leave the gang. They can step back

from activities to what is called a “calmado” (calmed down). These members are still

considered a part of the gang but not in active duty. The gang has to authorize a

calmado with religion being one of the only reasons for its acceptance.

• Attend meetings. These serve multiple purposes like collecting dues, talking strategy,

and comradery. They are also where the gang decides who lives and who dies. Meeting

14 are the place were every gang member is held accountable for their actions. Respect

for these meetings and the decisions is also a requirement.

• The beating and the mission. El brinco (the beating) and The Mission are the initiation

ceremonies for MS-13 and Barrio 18 respectively. Every member has to go through a

similar process of getting beaten up. Barrio 18 has one member count to 18 while four

other members do the beating (Dudley, 2015). While MS-13 allows a clica member to

determine the duration. Experiencing this ceremony is a requirement for becoming a

member.

Gang Violence

Violence is an integral aspect of gang life: it is a requirement to experience violence in order to become a full member and this expectation carries over to day-to-day activities.

Acts of violence are considered necessary to maintain a system of reciprocal aggressions with antagonists (Amaya and Martinez, 2015). These antagonists can be rival gangs, up and coming members challenging current leadership, state entities such as the police, and civilians. Gang violence can be split into the following categories: intra-gang issues, inter- gang disputes, confrontations with the state, or attacks on civilians.

Intra-gang violence is prominent because the violation of any of the previously mentioned gang rules is punishable under gang “law”. The punishment is determined in their meetings which tend to include alcohol and marijuana. These are no guidelines on what each violator’s punishment could be, this is up to the ranflero or palabrero in charge. Every violation within the gangs could be considered punishable by death. Physical aggressions are a common punishment for violations. Not claiming the gang, using drugs, wearing the wrong colors and other violations of this nature are punished by beat downs and a word of caution (Amaya and Martinez, 2015). The severity of the punishment can be based on the severity of the violation but it could also be based on the marero’s threat to current leadership. Given

15 high levels of raids and incarcerations there are constant challenges for leadership roles; if a member is violating a rule current leadership may choose to use this as a moment to reinforce authority. This also means that an up and coming member can use violence to establish himself as a figure of power. Intra-gang violence is a requirement.

Rivalries are the backbone of a gang. Having an enemy gives these groups a sense of purpose and something to fight over. Violence against the enemy is expected and rewarded within each gang. Fighting over territory should include aggressive behavior. There is an expectation that the response to an attack should always be more violent and should have a greater impact on the opposing group to solidify domination presence over territory. This has been a big cause of criminal violence in El Salvador where gangs operate in extremely close proximity to each other and have constant disputes over territories and passages. If a gang controls a street, members of the opposing gang cannot walk or drive by it. This problem has spilled over to gang member families making it difficult to transit through neighborhoods.

Violence against the state is manifested as violence against the police. They are seen as the executors of state orders and perpetrators of violence against the gangs. This became evident when the gangs agreed to a truce in 2012; when the truce ended, both MS-13 and

Barrio 18 had identified a common enemy. Police murders spiked, the response from the state was violent as well. The fight against the state has changed the dynamics of violence in El Salvador because for the first time it gave the gangs a chance to work together against a greater enemy.

Civil society is the recipient of spillover effects and experiences violence at multiple levels.

First hand violence from gangs in the form of extortion and harassment is common, becoming more violent when opposing gangs gain control of territories nearby. There has been an increase in the level of protection that civilians are willing to pay for in order to maintain their neighborhoods gang-free. Unfortunately this has been reserved for wealthier communities.

Disputes between gangs tend to affect entire communities and with the increased availability

16 of firearms, these have gotten worse for bystanders. Gangs are also engaging in violent acts against society with the intent to intimidate and control certain areas. They recruit for members from these communities and expect loyalty from the entire neighborhood. If this loyalty is not upheld at the gang standard then they respond with violence to instill fear and control.

Describing gangs and their criminal nature is critical for understanding the dynamics of violence in El Salvador. This is a small country almost completely urbanized leading to different networking connections. The socioeconomic standing of the country along with the policies implemented have made it the perfect terrain for gang growth and specialization.

This study addresses three issues that are affecting El Salvador that should be evaluated within the gang context before it continues and spreads to other countries with a large

MS-13 and Barrio 18 presence such as Honduras and Guatemala.

The second chapter of this research looks at the timing of policy implementation in

El Salvador. Using break point analysis I highlight how repressive policies in the country happen to be applied shortly before election periods in an attempt to sway votes in favor of political parties that used force against state enemies. Unfortunately, El Salvador has been a country where the government has not been able to maintain a monopoly of power; instead country leaders have used the perception of fear to highlight how criminal groups may be the true sources of violence and how they need to be stopped at any cost. Break point analysis highlights how mano dura policies have ultimately affected homicides. The implementation of repressive policies was favored by the media and yet again society accepted government’s use of punitive actions against the gangs. As political parties in El Salvador fought over power they manipulated the media and influenced people’s perceptions of violence to fit an agenda that promoted ”punitive populism”. Instilling fear in society has been a tactic used by both parties in order to gain and maintain power despite the obvious unintended consequences of these actions.

17 The third chapter evaluates criminal dynamics from a network perspective. Given that homicide has been has been considered a pandemic in the country, I use a method from the epidemiology world. It is a method that is better suited for evaluating the spread of a disease because it only considers its own impact and not external variables. This method steers away from traditional views of criminology such as nearest neighbor by looking at the impact that one municipality’s homicide rate can have on any other, not just those adjacent to it. Studies on Mexico’s drug trafficking networks have used this method to relate criminal activities beyond neighbors, considering the reach of cartels for sending messages and responses to their rivals. The method works well for the case of El Salvador because it takes into consideration gang responses beyond neighborhood crimes. The fourth chapter of this research consists on the creation of a gang ontology for event data generation on Mara Salvatrucha (MS-13) and Barrio 18 in El Salvador. This chapter seeks to identify the tactics the gangs use according to newspaper reports. It capitalizes on the detailed reporting mechanisms used in the country. In order to identify these spe- cific patterns I use event data analysis on newspaper stories from El Diario de Hoy to help identify coercive methods used by las maras. While homicide statistics are relatively re- liable and readily available, they do not distinguish between crimes and perpetrators, or identify gang-related violence. Despite pressing policy challenges, there is a considerable lack of gang-related violence data, even more so in Central America. Most accounts of gang violence rely on national homicide statistics, which do not differentiate between types of crimes. Although these are important, they are not adequate to understand the spatial dy- namics of gangs. This lack of information prevents local authorities, policy experts, and the international community from fully understanding the problem and implementing regionally specific policies to combat the issue. This research seeks to provide unique data on gang activities in El Salvador and their impact on violence across the country.

18 CHAPTER 2

IRON FIST FOR VOTES: THE EXCHANGE OF PUNITIVE POLICIES FOR

POLITICAL GAIN

Authors – Viveca Pav´onTercero

The Department of Public Policy and Political Economy, GR31

The University of Texas at Dallas

800 West Campbell Road

Richardson, Texas 75080-3021

19 Abstract A fundamental responsibility of the state is to maintain a monopoly on violence, unfortunately weak states can be challenged in this function by external forces. In an attempt to maintain control over a territory state leaders may chose policies that favor a public appearance of control. In other cases the state may even align itself with non-state actors in an attempt to regain control. Why would a state continue to implement policies that have failed at establishing a monopoly on violence? Is the perception of control as important as actual control for weak states? This study seeks to address these questions by analyzing the case of El Salvador, a country considered one of the deadliest in the world. Despite numerous failed attempts at controlling violence through punitive policies the country continues to enforce these actions and use them as platforms for future elections. Introduction In the years following the Chapultepec Peace Accords in 1992 El Salvador encountered an unexpected increase in homicide. The country had not healed from the war and was still experiencing division and ideological disputes. By the end of the decade homicide rates were dropping but this was not long lived. Before national elections in 2003 the government found a common enemy society could rally against, las maras. The fight against this newfound target provided common ground and a form of country unity. National media outlets embraced this fight against gangs and increased their coverage of anything related to crime reduction. Punitive policies against crime have been the norm ever since. High levels of perceived vi- olence have encouraged governments to continuously increase their punitive approach. Crim- inology theories of deterrence state that in order to decrease violence and crime the severity of the punishment should be increased (Vito et al., 2007). Central America has followed this school of thought and have created “zero tolerance” policies such as “Mano Dura” (Iron Fist). Guatemala and Honduras were the first countries to adopt these policies in 2001 and 2002 respectively. But it was the implementation by El Salvador that really characterized the repressive capabilities of such policies.

20 Was the crime in El Salvador that much greater than in its neighboring countries to require extremely punitive policies? What were the threats to society during the early

2000s that justified these laws? But more importantly, why have these policies continued throughout the years without much change in either legislation or actual improvement in crime? This paper seeks to uncover the political climate that initiated and gave perpetual power to mano dura and all other policies that have stemmed from it. By analyzing criminal trends we can better observe what the outcome of these policies has been. This paper adds to the current literature on gangs in El Salvador by studying all election periods since 2003 and the policies that have been enacted to deter crime by gangs. It also includes a study of homicide data that outlines changes in homicide trends, I associate these to previous policies for a better understanding of what is actually decreasing criminal violence in the country. Other studies have focused on qualitative accounts such as interviews and media reports providing a strong argument on what transpired in El Salvador. This paper adds a quantitative layer by allowing the data to indicate what years marked a significant change in homicide trends, then relating these trends to enacted policies. The paper proceeds as follows. The first section introduces the literature on mano dura related to elections and voting. The section also contains details on the current methods used to evaluate the relationship between the media and elections. Section two is an introduction to State

Power in El Salvador. This is followed by an explanation of the introduction of mano dura policies. I then present the case that political pressures affected how intensely punitive policies were enforced and present the data and method for relating policies to elections.

This is followed by an explanation of unintended consequences. The next section addresses how despite knowing the consequences punitive policies continue to be used by both ARENA and FMLN. The paper then addresses a few rehabilitation and reintegration movements and concludes with a discussion of the importance of timing for policy enforcement.

21 Literature Review

Mano Dura policies have been heavily studied from many angles and for several countries.

Studies on its punitive nature and it effects on society and criminal groups is quite vast for

Central America, the case of El Salvador is no different. The study of how political parties

and elections influenced mano dura is limited though. Wolf (2011, 2012, 2017) published a

book titled Mano Dura. The politics of gang control in El Salvador were she addresses a

connection between newspapers and the conservative party of El Salvador in order to instill

fear in society. Fear was meant to promote the use of punitive policies being proposed

by the right-wing party, Alianza Republicana Nacionalista (ARENA), yet opposed by the left Frente Farabundo Mart para la Liberacin Nacional (FMLN). Aside from her book Wolf has a few articles were she addresses the connection between the elites and politics in the country and how elite sentiment tends to dictate policy execution in exchange for votes.

Hume (2007) focuses on a historical narrative of the fragility of democracy in El Salvador and the shift towards authoritarian rule through fear. Hume seeks to understand why the government responded with punitive policies but instead explains the connection of election periods to policy launch. Authors such as Quintanilla Avalos (2017); Huhn et al. (2009) are also looking at this connection between policy introduction and elections but they do so by evaluating media coverage of gang activities. Huhn et al. study Costa Rica, El Salvador, and

Nicaragua. The authors trying to evaluate how newspapers in these countries address ’talk of crime’. Quintanilla on the other hand focuses only in El Salvador and is looking for a shift in publication of newspaper front pages in favor of ARENA proposed policies. Her study also makes a connection between the media and the elites and hoe they seek to affect societal sentiment with sensational stories about gangs. This study does not include any quantitative methods and focuses on newspaper story counts. Holland (2013) has a detailed study of mano dura and the role that elections and political pressures had on its implementation.

She bases her research on public opinion polls from Instituto Universitario de Opinin Pblica

22 (IUDOP) and Latin American Public Opinion Project (LAPOP) but narrows her research to

ARENA presidencies. Her approach is based on right-wing policy implementation under the assumption that left-wing parties would not implement repressive policies. Other authors and area specialists such as Cruz (2010, 2012); Martinez and Sanz (2012); Martinez et al.

(2013) also address this connection between the media, the elites, and political parties with the implication that they seek to gain political power by influencing society into believing gangs can only be addressed via punitive policies. Several of these articles are not directly about this relationship but they do address this connection.

State Power

The 1992 Peace Accords focused on demilitarization and democratization, this set the stage for El Salvador to become an electoral democracy. The accords facilitated the estab- lishment of FMLN as a new political party. ARENA, the political party in power at the time, and FMLN signed the Accords focusing on political and military reforms. Unfortu- nately, they failed at incorporating amendments that would improve poverty and inequal- ity. Instead, ARENA implemented neo-liberal economic reforms that aggravated inequality, particularly in urban settings (Van der Borgh, 2000; Richani, 2010; Montoya, 2012). San

Salvador experienced mass internal migration during the civil war as peasants moved seek- ing refuge for rural battlefields. After the war, migration continued due to the decrease in peasant economic opportunities as a result of neo-liberal policies. Rapid urban growth was a contributing factor for increases in crime over the next few years(Moodie, 2011; Garni and

Weyher, 2013; Quintanilla Avalos, 2017). There was also an increase of youth gangs in urban areas in the early 1990s but they were mainly informal groups concerned with local issues and only engaging in petty crime (Wolf, 2011).

A World Bank study in the late 1990s identified two different types of violence in El

Salvador: financially motivated and physical violence. Violence motivated by financial gain was limited to urban areas, the increase of this type of violence could be attributed to lack

23 of economic opportunities or the increase of street gang members engaged in petty crimes.

Physical violence against individuals on the other hand was predominant in rural areas and could be attributed to “delayed revenges” from the civil war (Cruz, 2010). This type of violence appeared to not have any criminal motivation and in many instances affected entire families. There was great concern that these killings could lead to political violence and another uprising. Further investigations revealed that while some of these killings had been committed by death squads others were simply private citizens settling unresolved issues from the war (Bejar, 1998). Revenge was a strong motive for the killings, families with the

financial means hired hit-men or former guerrillas to murder those they felt had wronged them. Revenge crimes were the first indication that the people of El Salvador had not reached peace but were instead settling old scores on their own. These types of crimes were heavily covered by the media, this influenced people’s perceptions of insecurity in the country.

Public opinion surveys recorded and increase in the population’s concern over security from one percent in 1991 to 40 percent by 1994 (Londoo and Guerrero, 2000).

Max Weber identified the monopoly of violence as a crucial component of a state’s ca- pacity. He defined the state as the entity that should oversee the use of violence and be the the only entity with the capability of using physical force. Charles Tilly also wrote about the importance of the state maintaining illegitimate violence in check and setting the guide- lines for the use of legitimate violence (Tilly, 1985). Following these descriptions of state monopoly of violence a government should not only be directly related with enforcing laws to limit violence but it should also have control over who else engages in violent acts (Cruz,

2010). After the establishment of a democratic government and establishing proper govern- ment institutions, El Salvador was not able to grant monopoly of violence to the state. The war ended due to international pressures in a battle that had reached a stalemate. This did not mean that ideological differences or political differences had been resolved. This lack of closure and recent experiences of battle allowed individual citizens the opportunity to seek

24 revenge and engage in criminal acts. The state could not control these situations, its lack of state power and monopoly of violence was quite evident.

Before the mass deportation in 1996, gangs in El Salvador were considered a social exhibit of belonging by the youth. Their baggy clothing and backwards baseball caps were recog- nizable yet a socially acceptable style. This concept of ‘gangsterism’ was almost promoted

(Valencia, 2014). The influx of gang deportees changed this perception as gang membership grew and clicas engaged in turf wars. Intra−gang disputes gained notoriety as MS-13 and

Barrio 18 took over most smaller clicas in the country and tagged streets and neighbor- hoods across the country. Maras became synonyms for crime and violence in the country

(Quintanilla Avalos, 2017). El Salvador is experiencing collective violence in response to a repressive state.

Mano Dura 2003

In an attempt to control criminal violence the government of Francisco Flores (1999-2004) implemented the first version of mano dura on July 23, 2003. Plan Mano Dura was a set of amendments to the penal code and judicial procedure code that would ideally remove gang members from the streets (Wolf, 2011). The plan also included a Ley Antimaras (anti- gang law) that authorized the arrest of potential gang members based on their appearance.

This legislation had three broad objectives: the arrest of suspected mareros, removal of graffiti used to mark gang territory, and allowed for large sweeps in search of gang members

(Wolf, 2012). Through mano dura the government criminalized youth gangs. The police were given greater power and the military was sent to the streets to assist in neighborhood sweeps. The police-military patrols resulted in mass arrests and limitations of civil rights

(Cruz and Carranza, 2006; Hume, 2007). National police reports detail that in the first two years of mano dura policies there where 30,934 arrests of suspected gang members (Aguilar and Miranda, 2006). It should be noted that this number includes multiple arrests of the same individuals. There were several reports of a suspected gang member being arrested

25 and detained for 48 hours only to be released due to lack of proof of gang affiliation. The same individual would be arrested in the near future. However, the over 30 thousand arrests where used as an indication of initial success of these punitive policies.

Since its origin Mano Dura has been highly publicized in El Salvador. Media outlets had been invited to the launch of policies when President Flores promoted the legislation from a neighborhood known for its gangs covered with graffiti on the walls. They continued to broadcast about the evils of gangs and were even included in raids so the evening news could cover the arrests of young men with tattoos (Quintanilla Avalos, 2017). The media and the government portrayed gangs as a major security threat (Peetz, 2008). Yet public opinion polls before the introduction of mano dura revealed that overall crime, poverty, and economic issues where the key problems society faced. A poll conducted by the Central

American University’s Polling Institute (IUDOP) in October 2003 uncovered gangs as a significant problem in society with 20.8% of respondents identifying it as a pressing issue.

Before punitive policies were introduced this statistic had not been greater than one per cent (Hume, 2007). This change in perception can be attributed to the constant publicity on raids and incarcerations.

Unintended Consequences

The forensic institute of El Salvador Instituto de Medicina Legal (IML) breaks homicide down by categories. In its 2009 report on the epidemiology of homicide it indicated that gangs were responsible for 4.3 percent of homicide in 2002. This number increased to 13.4 percent by 2005 (Molina Vaquerano, 2009). Even though these are relatively low percentages they do signal that crimes attributed to gang members became more prominent after the implementation of punitive policies.

The adverse effects of Mano Dura are also evident, and probably more significant, in incarceration rates. Figure 2.1 is a count of prisoners in El Salvador from 2004 to 2018.

There is an obvious increasing trend on the prison population in the country. The trend

26 appears to have a small dip between 2012 and 2014, this can be attributed to the gang

truce. Mano Dura empowered the police, encouraging them to arrest anyone that appeared guilty. Even though the judicial system tried to reduce the number of incarcerated young men without proof of crime the executive and the media criticized them for being “soft on crime”. The media further encouraged repressive policies by publicizing gang related activities instilling fear in society. Research on front page headlines of newspapers in El

Salvador show that between 1999 and 2002 there was one headline related to gangs; this number increased to 33 headlines between 2003 and 2004 (Quintanilla Avalos, 2017). The climate of fear helped promote Mano Dura and the successful policies of ARENA. Public

support led to more publicized sweeps and an increased number of incarcerations even when

the majority of these arrests resulted in suspected gang members being released after 48

hours.

Within the prison systems an unintended consequence emerged. In prison gangs where

given a safe space to meet and get to know members of clicas across the country, increasing

their overall network cohesion. Maras found their headquarters, they transformed their

operation systems and adopted the leadership structure that had been proven to work in

Los Angeles the previous decade. Inter gang disputes forced the government to segregate

prisons by gang affiliation (Cruz, 2010), this had two outcomes. First, “national leaders”

discovered that prison was the perfect place to have state granted protection from opposing

gangs and a place to establish their authority by granting protection and basic necessities

to those incoming gang members. Second, it encouraged incoming suspected members to

claim allegiance to one of the gangs in order to receive protection and rewards once inside. It

is during their time in prison that gang members received training and special instructions

to be executed once they were released (Wolf, 2011). Figure 2.2 is a count of incarcerated

gang members for 2003, 2006-2017. There is an steep increase on incarcerations starting in

2005. The number of incarcerated gang members was low in 2003 without much increase by

27 Figure 2.1. Prison Population Source: adapted from (DGCP, 2018; IUDOP, 2014, 2013)

2006. However, in 2007 this number increased making up 35% of the total prison population

(Cruz, 2010). One explanation for this increase in gang prisoners is Mano Dura. This policy relied on imprisonments. However, the policy did not provide adequate training to identify actual mareros and many young men without gang affiliation were imprisoned without proof.

This introduces the second possibility of why gang imprisonment increased by 2007 when young men kept being arrested and realized they needed some form of protection during their imprisonment. Many men chose to admit to belonging to a gang and begin their initiation process from within the prisons. Others simply realized they would continuously be arrested because of their physical appearance and they might as well reap the benefits of belonging to a gang (Martinez, 2017).

28 Figure 2.2. Incarcerated Gang Population Source: adapted from (IUDOP, 2014)

With a recidivism rate of over 90% for mareros the prison system became a right of pas- sage for young men joining a gang and turned street-gangs into more sophisticated criminal organizations (Lessing, 2014). Imprisoned Barrio 18 leader “El Viejo Lin” explained the effects of Mano Dura from the gang’s perspective: “Before [El Salvadors Mano Dura mass- incarceration policy] began it was different. We hadnt gotten to seeing things collectively.

The system has united us... like it or not, we cannot look at things individually, because they havent treated us individually, nor have they pursued or locked us up individually.” quoted in (Cruz, 2010, 393). Gangs in El Salvador responded to repression by coordinating and es- tablishing a hierarchy. Imprisoned leaders set new rules such as prohibiting gang tattoos that previously made them easy targets for incarceration, moved meeting off the streets and into safe houses, set specific roles for members of the clicas, and established la renta (the rent), prison-coordinates extortions (Aguilar and Carranza, 2008; Cruz, 2010; Savenije, 2009; Wolf,

29 2012, 2011; Quintanilla Avalos, 2017; Lessing, 2014). Gang recruitment also changed. High

levels of incarceration meant that new members had to be recruited to replace imprisoned

members. Neighborhoods began to experience forced recruitment of young men and violent

retaliations towards those who did not comply. Mareros on the street where expected to

provide for imprisoned members and their families which increased extortions and violence

(Cruz, 2012; Lessing, 2014).

Political Pressure

Mano Dura was introduced eight months before a presidential election. The ARENA government had been in office since before the signing of the Peace Accords in 1992 and had consolidate itself as a right-wing conservative party in favor of the country’s elite. ARENA had not been concerned with crime as a broad social issue before 2003 but increased mo- mentum by FMLN and economic concerns forced the party to reconsider its stance. The dollarization of the economy in 2001 was particularly detrimental to the lower and middle classes. Latinobarometro survey data indicate that by 2002 close to 60 percent of the country considered unemployment and poverty the biggest problems in El Salvador. Francisco Flores and the ARENA party where experiencing negative press and social discomfort. FMLN had been gaining momentum and had secured more votes than ARENA in the 2003 legislative elections (Holland, 2013). With a presidential election approaching ARENA enacted a plan that capitalized on citizen fear. Excessive media coverage of gruesome crimes further in- creased fear and helped ARENA gain popular support with its Iron Fist approach against gangs. Public fear of gang crime increased along with the view that ARENA was the only political party that was capable of decreasing gang activities. Any opposition to Mano Dura

was portrayed as siding with criminals by both the media and the government. Mano Dura

became ARENA’s winning strategy for the 2004 elections.

President Antonio Saca (2004-2009) from ARENA, began his presidency by institution-

alizing the legislation introduced by Francisco Flores. Because the ley antimaras included

30 military patrols for public safety it was considered an emergency law following the Peace Ac- cords of 1992. The Supreme Court deemed it unconstitutional two days before the law was to expire. However, a day after it was declared unconstitutional, an identical anti-gang law was accepted as a temporary law. Two months before this temporary provision expired Saca launched Super Mano Dura, an escalated version of the previous Plan (Quintanilla Avalos, 2017). Super Mano Dura lengthened prison times of gang members and increased military- police presence in the streets. The number of anti-gang task forces increased from 39 to 333 (Aguilar and Miranda, 2006). Punitive policies became the backbone for ARENA politics. They maintained presidential control through fear of crime and gang repression. Homicide was used as a driving force to encourage punitive policies but homicide data does not match the rhetoric of the time. In fact, the outcome of these policies was the opposite of what the government promised. To better understand homicide data I use structural break analysis to determine if there are any breakpoints around the time Mano Dura policies were implemented, in order to asses the impact of these policies. This model will point-out any significant breaks in the time series data. Government officials in El Salvador expected that Mano Dura would decrease homicide, making the hypothesis: the implementation of punitive policies will lead to a decrease in homicide. Breakpoint analysis removes the possibility of introducing any pre-determined expecta- tion into the data. Its focus is to allow the time series to to speak for itself. Following the Bai and Perron (1998) approach, I test for unknown breakpoints on monthly homicide rate per 100,000 from 1995 to 2017 by using Bayesian Information Criterion (BIC) 1. The occurrence with the lowest BIC is considered the optimal model with the ideal number of breakpoints. 2 Figure 2.3 is the optimal model with four break points: the first in April of

1National homicide data for El Salvador extracted from National Police sources, DIGESTYC, and The World Bank data bank from UNODC sources. Data is defined as Intentional Homicides

2BIC and RSS plots can be found in Appendix A.1

31 Figure 2.3. Monthly Homicide 2002-2017 with significant structural breaks

2004, the second in December of 2008, third in February of 2012 and the last in June of 2014.

The structural break of 2004 shows the start of a increase in homicide a few months after

Mano Dura was implemented. This was a significant break from homicide in 2003 when th country had a low homicide rate of 37.2 (Bank, 2018). The data does not find a significant structural break before 2003. This is a contradicting narrative of what President Flores and the media portrayed in 2003. Instead what we can observe is an increase in homicide after punitive policies were implemented. Because the 2004 structural break is after the Mano

Dura policies where implemented the data suggests that punitive policies did have an effect on homicide. Given that there is an increase in homicide after the policies where in place there is an indication that Mano Dura and Super Mano Dura had an opposite effect than what was intended. In fact, every structural break has the opposite effect that was intended

32 and they all align with electoral periods. Table 2.1 lists the electoral months and years for

El Salvador staring in 1999 to 2015.

Table 2.1. Elections in El Salvador Election Type Year Month Presidential 1999 March Legislative, C.A. Parliament and Local 2000 March Legislative and Local 2003 March Presidential 2004 March Presidential 2009 March Legislative and Local 2012 March Presidential 2014 February Presidential 2014 March Legislative and Local 2015 March Source: Author’s creation adapted from (TSE, 2018)

The second structural break in 2009 is also followed by an increase in homicide. This

time period also happens to match an election year. Interestingly though during this election

period ARENA lost and for the first time in El Salvador’s history FMLN had won the

presidency. The third structural break is in 2012, which is the year the truce between the

gangs was established. This truce helps explain the big drop in homicide for 2012. The last

significant structural break occurs in 2014, yet again an election year, and also followed by

an sharp increase in homicide.

Winning Elections

President Francisco Flores introduced a policy that would distract society from their

most pressing issues. A weak economy was not something ARENA could fix before the

elections but they could launch a campaign where they would be seen as solving an important

issue. Unfortunately the homicide data and the breakpoint analysis presented in this paper

contradict the campaign’s claims. First of all, there was no significant gang issue before

Mano Dura. Also, if we use homicide as a proxy for violent crimes attributed to gangs, we can see that its increase occurred after repressive policies where in place. Mano Dura as a

33 citizen security measure was a failure however it proved to be a politically profitable measure

(Cruz, 2004; Wolf, 2011). There are two critical reasons for this “success”: citizen fears and

the role of the media. These two factors are interconnected.

The use of fear as a campaign tool is not new in El Salvador where fear of communism

helped shape the civil war and fear of political violence shaped elections. ARENA politicians

expected this concern to persist and weaken FMLN, but in 2003 the leftist party was gaining

momentum threatening the presidency in the upcoming elections. ARENA was a conser-

vative party that catered to the elites but had won over rural voted in previous elections

(Holland, 2013) given that the war was more damaging in rural areas and FMLN leaders

had been war actors in those regions. Economic policies in the early 2000s favored business

elites and urban sectors, leaving rural voters dissatisfied with Flores’ presidency. ARENA

needed a way to attract both affluent elites and poor rural voters. Mareros became the new

threat for society and the fear the party would use in its campaign. Unique wardrobes, hand

signals, tattoos, and petty criminal activities made it easy for the government to label gangs

as dangerous groups. A newfound enemy of all of society played on people’s fear of criminal

violence and helped rally votes in favor of the conservative party. Mano Dura was introduced

in a perfectly staged situation where President Flores was surrounded by police and military

leaders in a neighborhood with a high crime rate. That same night military-police raids were

broadcast by the media showcasing shirtless young men with tattoos (Quintanilla Avalos,

2017). This was the embodiment of mareros, what society should be afraid of, the new

’stereotype of threat’ (Hume, 2007). It was also a problem ARENA was prepared to fix.

The media sensationalized these events. They repeated Flores’ rhetoric that gangs were

responsible for the majority of homicides and the only way to stop them was to implement

repressive policies. The role of the media was not entirely surprising, after all sensationalist

stories sell newspapers in Latin America, particularly because La Prensa Grafica and El

Diario de Hoy, the top selling newspapers in El Salvador, are owned by families with business

34 ties to ARENA. The country’s journalistic environment has been affected by this allegiance to

the extent of censoring journalists that steer away from the newspapers ideology and political

stance (Wolf, 2011; Freedman, 2012). The radio environment was even more partisan given

that presidential candidate, later president, Antonio Saca owned 7 radio stations in 2003

(Freedman, 2012). ARENA also had strong ties with television channels. These three

mediums of information promoted Mano Dura policies and increased their coverage of gang related crimes. The media was crucial for enhancing citizen fears. News coverage included reported crimes but with little oversight on journalistic quality more reports were focused on journalistic opinions. Stories of funerals and images of grieving mothers became the norm.

The media in El Salvador found a way to increase their coverage on gang related crimes by blaming the gangs for any crime or by including opinions on the need for repressive policies

(Quintanilla Avalos, 2017).

Public opinion polls suggest that society changed its view on what the most pressing issue in the country was. Polls from LAPOP and IUDOP both indicate that crime became the most pressing issue after the implementation of punitive policies in 2003 (Holland, 2013).

This was a clear indication that media coverage had heightened the gang issue and increased the importance of stricter security measures. This made the use of media for political purposes an extremely significant tool during the early stages of Mano Dura in El Salvador.

ARENA won the 2004 elections. President Saca launched Super Mano Dura early on in his presidency (August 9th, 2004) and maintained punitive policies as a focal point of his presidency. He relied on media coverage of arrests to increase public opinion of ARENA and his presidency and continued to use negative coverage of gangs to promote further repression.

A 2004 report in El Diario de Hoy titled “Al Qaeda with gangs?” addressed gang members as terrorists with links to Al Qaeda. Amid controversy over this claim President Saca stated that he did not have even information on the suspected links with terrorism but would not dismiss the idea that the maras had a connection with terrorist groups (Quintanilla Avalos,

35 2017). Public concern increased, helping label gangs as terrorist organizations when the Anti- terrorism Law went into effect in October of 2006 (Lohmuller, 2015). This law increased the number of years arrested gang members would spend in prison if they were found guilty of crimes against individuals. It has been categorized as a new form of government repression at a time when homicides and victimization was increasing.

ARENA’s ‘punitive populism’

Mano Dura was intended as a policy approach implemented for political gain. ARENA’s use of policy with the expectation of winning votes disregarding the outcome of the pol- icy is a form of populism. In their study on the Supreme Court of El Salvador Martinez

Barahona and Linares Lejarraga (2011) explain that ‘punitive populism’ involves the ex- ploitation of misinformed opinion in the pursuit of electoral advantage in the sense that many manifestations of penal populism are premised on assessments of public opinion which are oversimplifications or incomplete in some way” (Martinez Barahona and Linares Lejar- raga, 2011, 53). In this context the Flores and Saca administrations acted as if they were meeting society’s demands regarding security when their own administrations created the exaggerated notion that gangs were the main culprits of crime. This is not to say that gangs were not engaged in criminal activities but homicide data shows that this was not a pressing issue at the time. The policies implemented where opportunistic playing off of citizen fear that was constantly fed by the media.

FMLN Presidency

The 2009 election period was difficult for ARENA, their policies had backfired and crime was a bigger issue now a bigger issue. Unfortunately gangs had organized and established economic pressures on society in the form of extortions. Citizens had become actual victims of crime and were no longer experiencing just the perception of it. Incarceration rates were at their highest levels to date and the prison system was extremely overpopulated. The use of the media had helped them in the previous election but now it was a constant reminder

36 of policy failures. Continuing with the strategy that helped them win the previous election

ARENA selected Rodrigo Avila, a two-time police chief as its candidate (Holland, 2013).

FMLN, on the other hand, steered away from punitive policies. They based their campaign

on a two-sided approach: first FMLN developed a bipartisan security plan that the Saca

administration shelved. FMLN used the security plan as an opportunity to create relation-

ships with ARENA votes. Second, they campaigned proposing crime control programs that

targeted the roots of violence including socioeconomic issues. FMLN’s candidate Mauricio

Funes was liked by businessmen and his more strategic approach to citizen security was

well received. took office June 1st 2009 (Holland, 2013). Even though he

implemented his own crime prevention policies he did not overturn Mano Dura policies.

Homicide rates continued to increase during Funes’ presidency. Legislative elections were scheduled for March of 2012 and FMLN needed some leverage. That same month the Minister of Security David Munguia Payes held secret negotiations with Mara Salvatrucha and Barrio

18 leaders, all of whom where in prison (Martinez, 2017). He brokered a truce between the

gangs where they agreed to stop killings in exchange of transferring gang leaders to lower

security prisons where they could get family visits and have easier control over the gangs on

the streets. There was an immediate drop in homicides, coincidentally on the weekend of

the elections. The Minister of Security claimed this drop was due to stricter military sweeps

(Valencia, 2012). A few days later on-line newspaper El Faro revealed that the drop in

homicides was due to the truce. The government denied knowing about the truce and being

involved in it (Thale et al., 2013). The uncertainty around this agreement was evident and

the mass media began calling it “the so-called gang truce”. Even though homicides where

dropping the truce was not a welcomed negotiation, the media highlighted the fact that if

the government negotiated with gangs they were gaining political power. La Prensa Grafica

had a survey following the truce showing that 55 percent of the population disproved of it

(Rubio-Fabian, 2013). Munguia Payes eventually acknowledged his involvement in the truce

37 but the Funes administration did not. Regardless the government benefited from the sharp decrease in homicide. The murder rate of El Salvador in 2011 was 70.1, in 2012 this figure dropped to 41.2, by 2013 the murder rate was 39.4 (Martinez, 2017). Despite this significant drop the gang truce did not have public support. A 2013 IUDOP survey revealed that 83 percent of the population had little or no trust in the truce (IUDOP, 2013).

2014 was an election year. ARENA experienced internal disputes after it lost the 2009 elections. This resulted in the splitting of the party and the creation of a new one: Gran

Alianza por la Unidad Nacional (GANA). FMLN continued to deny its direct involvement with the truce and realizing that society wanted an authoritarian response to crime increased its repressive measures. Police operations and mass arrests became easy publicity stunts.

For the sake of the party Funes went back on his campaign promises and used a Mano Dura approach to fight crime (Martinez, 2017). FMLN candidate Salvador Sanchez Ceren won the presidency by a slim margin. ARENA claimed fraud. Both political parties rejected the truce during their campaigns and vowed to not engage in deals with mareros. The increase of repressive measures and the rejection of the truce by FMLN leaders weakened the deal made with gang leaders, by the end of the year the truce between the gangs was dissolved.

Homicides increased at an unexpected rate, reaching its highest level since the civil war in

2015 with a rate of 103.6 (Martinez, 2017). Figure 2.3 is the time series of homicides with four structural breaks. The first three have been previously discussed but the last breakpoint was identified in 2014. The dissolution of the truce marked a new era for both gangs and the government. Gangs realized that they had more power in society when there were killings, people never liked the idea of inclusion of the gangs and rejected them as members of civil society. When there was an increase in violence though, society respected the gangs. The government did not want to make deals with the gangs even if this meant murder rates would increase.

Back to punitive policies 2015

38 Faced with public pressure due to the increase in violence, the government of Sal- vador Sanchez Ceren unleashed police and military forces onto marginalized neighborhoods.

Clashes between the police and gangs where violent and almost always ended with several members of both sides killed. Gangs began to target police members and their families.

The police responded by engaging in clandestine raids. In January of 2015, the chief of the PNC Mauricio Ramrez Landaverde instructed the police to “shoot delinquents without fear” (Ramirez Landaverde, 2015). In media conferences Ramirez called upon the police to use their fire arms and feel confident that the institution would back them. As of 2017 the

Procuraduria para la Defensa de los Derechos Humanos (National Prosecutors Office for the

Defense of Human Rights) of El Salvador has processed 37 cases of extra-judicial executions

(Martinez, 2017). Most of these cases have been identified as killings of mareros that had already surrendered. Surprisingly there have also been reports of police killings of gang family members.

Clashes between the police and gang members increased. Targeted killings from both sides became every day news. The war between the state and the gangs had reached its peak. In August of 2015 there was one murder per hour in El Salvador (Watts, 2015).

49 police officers had been murdered in those nine months of the year (Cruz et al., 2017).

The government responded by drawing upon the 2006 Anti-terrorism Law, issuing warrants against 300 gang members accused of “terrorist acts”. El Salvador’s Attorney General,

Luis Martinez, accused gangs of terrorizing the population and being responsible for at- tacks against government officials and police officers. Gangs were also responsible for the displacement of at least 533 individuals (Albadejo, 2016). Luis Martinez stated that these accusations along with every day crimes attributed to crimes where a clear indication that gangs where trying to destabilize the government, justifying the use of the Anti-terrorism

Law against them (Lohmuller, 2015). Repressive policies have been in place since.

Rehabilitation & Reintegration

39 Given that gangs were not a pressing issue in El Salvador in the late 1990s and early

2000s government efforts for rehabilitation or reintegration did not exist. There were a

handful of civil society funded rehabilitation programs that were underfunded and did not

play much of a role in the media (Aguilar and Miranda, 2006). In 2004, a year after the

implementation of punitive Mano Dura, President Antion Saca (ARENA) faced criticism by human rights groups; he responded by launching Mano Amiga (Friendly Hand) and Mano

Extendida (Extended Hand). Mano Amiga focused on early prevention programs while Mano

Extendida was based on reintegration of gang members into society. Both of these programs

were underfunded, poorly executed, and faced social criticism (Wolf, 2011), they did not

have an impact in rehabilitation or reintegration. In fact, rehabilitation policies where never

considered a “serious step toward less repressive policies” (Peetz, 2011, 1485).

When President Mauricio Funes (FMLN) took office in 2009 he was determined to im-

plement preventative policies. He began by establishing investigative policing and prison-

based rehabilitation programs (Wolf, 2011). Government funding for prevention programs in-

creased from 1 percent to 14 percent of the budget for the Ministry of Defense (Ribando Seelke,

2011). These steps were the first indication of the government moving in favor of rehabili-

tation and away from punitive policies. The media portrayed Funes’ approach as “soft” on

criminals and the public did not seem to agree with the government’s new policies. This

shift towards rehabilitation was short lived, after the attack on a public transportation bus

that ended with the death of 17 individuals was attributed to gangs, President Funes en-

forced gang criminalization. This response was welcomed by society. The government was

yet again enforcing Mano Dura policies. However, the 2012 gang truce included government

promises for reintegration and rehabilitation programs along with job training and educa-

tion (Martinez and Sanz, 2012). Once again FMLN was caught in the controversial stance

of cooperating and being “soft” on criminals. When the truce process fell apart in 2014

so did the rehabilitation promises. Salvador Sanchez Ceren ran his presidential campaign

40 against the truce but with security reforms which included rehabilitation programs. These materialized as El Salvador Seguro in 2015. This is a collaborative program including edu- cation, job training, rehabilitation and reintegration efforts at local and municipal levels. It also included community policing. Unfortunately, the program did not receive appropriate funding and has been implemented at its full capacity. El Salvador Seguro has been difficult to promote given the attacks against the police attributed to gangs. Rehabilitation and reintegration policies have been scarce and their implementation has not been well funded or accepted by society. El Salvador has not had the opportunity of testing alternatives to repression. A lack of public support for rehabilitation policies makes them unlikely future approaches. Discussion One of the key features of a democratic government is the ability to maintain monopoly of power. But when a state is incapable of being the sole agency overseeing legitimate uses of violence it must go through extreme measures to regain this control. In the case of El Salvador the government used fear tactics to gain votes in favor of implementing policies that resulted in the force that took the monopoly of power from the government. By using punitive policies before they were required the government empowered gangs and helped them become more violent. In this study I have gone through the history of Mano Dura and have showed how its implementation came at a time when gangs where not the main source of concern for society. In fact, homicide was on a decreasing trend. Break point analysis identified 2004 as the first year when homicide shifted from its decreasing pattern to an increasing one. This shift happens to occur the year following Iron Fist policies meant to decrease crime. The change in homicide trends can assist in relating how policies implemented before an election period serve the purpose of winning the election for the party in power but it was detrimental to the rest of society. In 2003, ARENA used its influence as a right-wing party in favor of democracy and institutions to promote repressive measures against an enemy most of society could identify.

41 Table 2.2. Mano Dura policies President Party Policy Francisco Flores ARENA Mano Dura (2003) (1999 - 2004) Ley Antimaras (2003)

Antonio Saca ARENA Mano Dura (2004) (2004 - 2009) Mano Amiga (2005) Mano Extendida (2005) Super Mano Dura (2006)

Mauricio Funes FMLN Anti-Gang Law (2010) (2009 - 2014) Gang Prohibition Act (2012)

Salvador Sanchez Ceren FMLN Mano Dura (2015) (2014 - ) Police Impunity (2015) Source: Author’s creation adapted from (Wolf, 2012; Peetz, 2011; Does, 2013)

The media backed the party’s proposal by increasing their coverage of gang related crimes. This became an easy task for newspapers and other media outlets given that gangs identified themselves with tattoos and hand gestures that intimidated non-members. The media played a significant role in influencing people into believing ARENA was the only party that could control the newly found gang problem. As political parties in El Salvador fought over power they manipulated the media and influenced people’s perceptions of violence to fit an agenda that promoted “punitive populism”. The same tactics were followed during the next presidential campaign. FMLN came into power in 2009 and there was an expectation that policies would shift to a rehabilitation approach, unfortunately this was unpopular and Mano Dura type approaches were used again. Back-door meetings and agreements led to the truce between the gangs and even though the government never admitted its participation this agreement has been the only program to reduce homicide. Break point analysis identifies 2014, the year the truce broke down, as the fourth break in the trend in homicide. Yet again breakpoint analysis identifies a point in time where homicide starts increasing. This is also the year the government allowed the police to respond to gangs without concern over

42 punishment and without due process. Table 2.2 lists how every presidency in El Salvador has been involved in the sanctioning of punitive policies regardless of political affiliation. Instilling fear in society has been a tactic used by both parties in order to gain and maintain power despite the obvious unintended consequences of these actions. It was clearly evident that gangs in the country found a space to organize and grow while in prison. Repressive policies may have had public support and may have been useful for winning votes but they have also been useful for giving gangs a platform for control. It is evident that even though the data shows that the only time there was a significant decrease in homicide was when the gangs engaged in a truce, the leaders of El Salvador are not interested in cooperating with the gangs. Instead it is more important to maintain public perception of power by implementing punitive policies that sell newspapers with sensationalist headlines over actual facts.

43 CHAPTER 3

THE IMPACT OF GANG NETWORKS

ON HOMICIDES IN EL SALVADOR

Authors – Viveca Pav´onTercero

The Department of Public Policy and Political Economy, GR31

The University of Texas at Dallas

800 West Campbell Road

Richardson, Texas 75080-3021

44 Abstract

The Impact of Gang Networks on Homicides in El Salvador In 2015, the homicide rate in El Salvador was 104 deaths per 100,000 the highest of any country in the world. While other Latin American countries, notably Mexico, experience similarly high rates due to narco violence, the spread of violence in El Salvador seems to be attributed to gang proliferation.

In fact, as of August 2015, the Supreme Court of El Salvador declared gangs are terrorist organizations. In this work, I analyze officially reported homicides by date and location to show gang violence contagion over space and time. The goal of this article is to see if gang violence is localized or it is coordinated in a national gang network. Based on correlation between municipalities and the change in homicide preliminary findings suggest that El

Salvadoran gangs are not only interconnected, but their interdependence and coordination is the driving cause of the increase in the national homicide rate. Contagion is also facilitated by the relative lack of rural areas.

Introduction

The municipality of Mejicanos in El Salvador is one of the most disputed territories in the country. From a relatively low homicide count of 10 in 2002, homicides peaked in 2009 at

116 for the year, then decreased to 92 in 2013 (Carcach, 2015). Mejicanos is predominantly controlled by the Pandilla 18, but the northern region of the municipality is controlled by rival mara MS-13. This co-habitation of gangs has led to multiplying homicides and crime instability affected by local conflict and external disputes. How can we explain the complex dynamics of criminal violence in Central America?

Mejicanos is just one example of the escalation of criminal violence in Central America that has escalated in the last decade making it the deadliest region in the world. The UNODC reports that the level of violence in this region is higher than in any single nation, including those at war. In 2012, Central America had an annual homicide count of 17,672 people, with 15,791 of these casualties in El Salvador, Guatemala, and Honduras, a country group

45 otherwise known as the Northern Triangle (UNODC, 2013, 2014). This staggering number is far beyond even the average civil war death toll of 10,500 (Lacina, 2006). Responding to this violence is the top policy priority in the region.

This paper uses homicide data for El Salvador, a country were killings have increased

70 percent in 2015 reaching a homicide rate of 104 people per 100,000 (Watts, 2015). This increase in homicide makes El Salvador the deadliest country in the world. This research seeks to introduce a new way of analyzing homicide patterns by evaluating the correlation of within-municipality differences in homicide with all other municipalities.

Previous criminal violence analysis sought to understand, describe, and even help pre- dict the dynamics and spread of conflict. Specifically, studies have focused on drug-related crime or gang-related violence networks through literature based approaches (?), diffusion equations (Pitcher et al., 1978), reaction-diffusion equations (Zeoli et al., 2014), and through mixed methods incorporating geography and historical accounts of conflict (Papachristos,

2009; Papachristos et al., 2013). Unlike many other studies that evaluate specific gang lead- ership roles and individual gang member activity, this study focuses on overall gang activities within municipalities of the country. Unfortunately, documented data on gang activity, or drug-related criminal activity, in the country is sparse.

This paper employs Social Network Analysis (SNA) to better understand the dynamics of criminal networks. I use this method because it calculates and visualizes changes in net- work members characteristics and detects active patterns of criminal activity by measuring correlation of municipality dyads. The visualization of these dynamic measures helps us better understand changes in relationships and the overall structure of criminal networks in

El Salvador. The paper proceeds as follows. The first section introduces the literature on organized criminal violence and relevant policies as well as current measures for countries with similar criminal activities like Mexico. In section two I introduce the expectation of the analysis and a short case study with expert opinion to help validate the prospects of the

46 paper. The third section describes the data. The following section introduces the methods used and the visualized results. The last section is a discussion of the results and a conclusion on the analysis.

Literature Review

Much of the literature on drug trafficking organizations (DTOs) focuses on Mexico or

Colombia. But recent reports suggest that El Salvador, Guatemala, and Honduras - the

“Northern Triangle – are at least as affected, if not more so, by drug-related violence. This geographical group differs from the rest of Central America by its higher levels of violence and criminal activity. With 2012 homicide rates of 90.4, 41.2, and 39.9 per 100,000 in Honduras,

El Salvador, and Guatemala respectively, the Northern Triangle suffers from the highest homicide rates in the region and the world. Gangs and DTOs are responsible for most of the bloodshed (Demombynes, 2011). Drug trafficking is commonplace in the Northern Triangle, with about 90 percent of the cocaine entering the United States (U.S.) from South America passing though the Central American-Mexico Corridors, and 80 percent of this amount first entering a Central American country before being smuggled into Mexico (INCSR, 2014).

Figure 3.1 is an annual sum of homicide by country in Central America and Mexico; it also includes points of DTO-focused policy intervention.

Qualitative accounts indicate an emerging relationship between the Maras and the ever- expanding DTOs, especially into geographically vulnerable countries like those in the North- ern Triangle. Area specialists such as Bagley (2013), Breve-Travieso (2007), Demombynes

(2011), Dudley (2010, 2011), Ribando (2005), as well as Torre and Martin (2011), address the connection between gangs and organized crime; yet, there are no empirical studies ex- ploring this connection as a joint problem. The data used in this paper does not differentiate between crimes committed by either group; instead, this research focuses on changes in ag- gregate homicide counts as the measurement of the impact both criminal groups may have on El Salvador. Figure 3.2 shows El Salvador’s homicide counts from 2005 to 2013.

47 Figure 3.1. Homicide in Mexico and Central America Author creation with Source: (UNODC, 2013) data

Organized Crime in El Salvador Organized crime is defined as a stable and hierarchical group of criminals with the ability to use violence for control of illegal markets in order to obtain economic rent (Reuter, 2014). A DTO is defined by the UNODC as a criminal organization with the objective of growing, manufacturing, transporting, distributing, and marketing substances questionable under drug prohibition law. Although the primary ob- jective of DTOs is to make money through the transfer of illicit drugs, they are also open to activities such as extortion, firearm smuggling and even human trafficking, kidnappings, and money laundering (UNODC, 2013). Similarly, a Mara is defined by the UNODC as a street gang with territorial control. In the case of the El Salvador, members of these groups tend to self-identify with tattoos. Given that there is an overlap in activities between these

48 Figure 3.2. Homicide in El Salvador 2005-2013 Author creation with data from Carlos Carcach groups, this paper addresses all of their activities as organized crime without perpetrator distinction.

Homicide counts in El Salvador have fluctuated through time, as shown in Figure 3.2.

National-level data, for the study time period, tally homicide at a high of 4,360 in 2011 and a low of 2,245 in 2012. Despite this drop, El Salvador has a very high homicide count in comparison to other countries; it is eight times deadlier than the United States for example

(Hope, 2014). But focusing on a national-level analysis detracts attention from the meaning- ful variation in lower units of analysis within the country (Storeygard et al., 2008). Figure

3.3 shows the 2013 homicide rate across all 14 departments (equivalent to U.S. states) in El

Salvador. The darker shades indicate a higher homicide rate and we can see a concentration in the middle portion of the country for the departments of Cabaas, Cuscatln, and La Paz.

49 Figure 3.3. 2013 Homicide Rate by Department Source: (Ingram and Curtis, 2014)

The rest of the country appears to have an even distribution of rates between 15 and 50 homicides per 100,000 people.

Going even deeper sub nationally, Figure 3.4 maps homicide across the 262 municipalities

(equivalent to U.S. counties) in El Salvador. The first obvious difference is the difference in scale: homicide rates at this unit of analysis are much higher, reaching levels of 153 per

100,000. This indicates greater variation in rates at smaller geographic levels where higher homicide rates are no longer concentrated in the center of the country. In this map there is a clear distribution of higher homicide rates along the Pacific coast, a feature that is not evident when evaluating data at the department level. It is because of the richer variation that I choose to analyze homicide at the municipal level.

50 Figure 3.4. 2013 Homicide Rate by Municipality Source: (Ingram and Curtis, 2014)

Despite multiple studies focusing on Mexico, scholars rarely study criminal violence in El

Salvador even though the country is constantly ranked as one of the deadliest countries in the world. Even though the circumstances for crime are different between the two countries, the spread of drug trafficking related violence appears to have similarities that we can infer from.

Researchers like Dell (2015), Osorio (2013); Osorio and Reyes (2016), Carcach (2008, 2015),

Alvarado (2011), Ingram and Curtis (2014) study the relationship between drug trafficking related crime and geography and how drug trafficking is a crime that needs to be observed at micro units such as municipalities instead of aggregate accounts. Dell (2015) and Osorio and Reyes (2016) focus on Mexico and how municipalities with known DTO presence have noticeable criminal activity. Dell (2015) finds that when elections affect a municipality there are shifts in the routes are used for trafficking, later becoming territories with larger levels of

51 criminal activity. Osorio and Reyes (2016) concludes that drug trafficking related violence is spatially auto-correlated, meaning that crime activities tend to cluster around each other.

Focusing on El Salvador, Carcach (2008, 2015) , Alvarado (2011), and Ingram and Curtis

(2014) identify clusters of homicide which are also regions known for their gang and, more recently, trafficking presence. Carcach (2015) concludes that we can identify spatial patterns and that crime risk in El Salvador is concentrated around specific geographic corridors and highways that have remained stable through time. Alvarado (2011) finds auto-correlation of homicide, runs a hot-spot analysis and concludes by pointing out vulnerable municipalities.

Ingram and Curtis (2014) find that using municipality-level data is a necessity when studying homicide in El Salvador and that aggregated homicide by type is spatially clustered. All of these researchers highlight the importance of drug-transit routes and territorial conquests.

Hypothesis Many of the previously mentioned studies focus on network analysis of drug- related violence or gang-related violence separately. Others focus on the spatial relationships of violent crime by following the nearest neighbor theory that crime will spill-over to adjoining locations, thus resulting in geographic clusters. This paper seeks to analyze criminal violence in El Salvador by evaluating transitions and relationships of homicide at the municipality level, hypothesizing that violence in a given municipality is positively related to violence in other areas of the country not only to adjacent municipalities. Prior literature identifies criminal violence hot spots but does not evaluate a connection between these areas. By implementing SNA methods, I expect to find correlation between some municipalities far apart in distance from each other because of geographic components and group rivalries.

Smuggling routes are not new in El Salvador, but policy and relations alter the importance of a specific route and potential fights over the territory. This is evident in cases where smuggling corridors shift in accordance with police intervention or trafficking directionality, very much like a balloon effect. I also expect to see homicide counts affected by gang relations, such as the introduction of a Peace Treaty in 2012 and the punitive nature of gangs where

52 responses are not limited to geographic centers, but are instead based on inflicting higher

retaliatory damage.

Research Design

This study uses SNA measures to connect municipalities with correlated homicide pat-

terns. Studies such as Krebs (2001); McAndrew (1999); Sparrow (1991) use similar methods

and found significant results for understanding the structure of criminal networks. Likewise,

? use networking to identify Mexican municipalities with linked drug-trafficking related

homicides with monthly time series data to understand the correlation in homicides within

municipalities controlled by cartels. I use a comparable approach to Espinal-Enriquez &

Larralde by evaluating similar patterns of homicide, but I extend the data to include all

homicides. I focus on annual data from 2005 to 2013 instead of monthly reports due to data

limitations.

The information used for this project was made available by Dr. Carlos Carcach of the

Escuela Superior de Economia y Negocios (ESEN) in El Salvador. Unfortunately, there is not

a single source of information at the municipal level. Carcach pieced together homicide values

by collecting data from Attorney Generals Department or Fiscalia General de la Republica

de El Salvador; 1995 - 2002 from El Salvador’s Institute of Forensic Medicine or Instituto

de Medicina Legal (1995 - 2002); and 2009 - 2014 from National Civilian Police or Policia

Nacional Civil. Population data came from the National Direction of Statistics and censuses or Direccin General de Estadstica y Censos (DIGESTYC) (Carcach, 2015). This data at national level per year is comparable to the homicide counts reported by the UNODC, as seen in Table 3.1.

Empirical Tests Intelligence agencies use networks to describe criminal organizations be- cause their offenses tend to rely on interrelated actions (Xu et al., 2004). These organizations are in constant flux with new members, role changes, power alterations, and clique creation.

Network analysis is useful in reflecting these changes. The structure of organized criminal

53 Table 3.1. Data Comparison Year Carcach Data UNODC 2005 3,701 3,778 2006 2,916 3,928 2007 3,484 3,497 2008 3,179 3,179 2009 4,382 4,382 2010 4,004 3,987 2011 4,360 4,371 2012 2,245 2,594 2013 2,513 2,513 Source: adapted from (UNODC, 2013) and Carcach data networks can be examined at three levels: node (municipality), link (correlated municipal- ities), and overall network. In this study I evaluate the impact of conflict between groups, which is inferred in the changes in homicide counts in individual municipalities over time.

First, I construct a correlation network between municipality pairs by calculating a Single

Datum Correlation Coefficient (SDCC):

2 [(Xi)(Yi) − (µi )] SDCC(xiyi) = (3.1) Nσi Using SDCC I calculate the joint variation in the change in the number of homicides every year with respect to the national homicide for a given year, between every pair of municipalities. This municipal variation is normalized by the country-wide variance of the given year multiplied by the number of municipalities in El Salvador. SDCC is calculated between every pair of municipalities for every year of available data (2005-2013), resulting in one SDCC network per year. Using this method I am able to evaluate the correlation that the change in homicide has had between every municipality from one year to the next. The

SDCC formula is run individually for ever municipality dyad where X i and Y i represent the change in homicide count for each municipality during year i with respect to the previous year; µi is average homicides over all of the municipalities during year i, while σi is the

54 variance for all of the municipalities for year i. N is the number of municipalities in the

country. I then proceed to connect the municipalities with an SDCC above a threshold θ.

These connections become the link of the network highlighting those nodes where there has

been a concurrent change in homicides.

Theta

The homicide network nodes are the municipalities affected interrelated changes in homi-

cide. Two municipalities (nodes) X and Y are linked if they satisfy

SDCC(xiyi) > θ (3.2)

where θ is a correlation threshold. The magnitude of θ represents how the SDCC was

normalized. Following SNA research from Espinal-Enriquez and Larralde (2015) a θ of 0.001 is used in this model as the acceptance threshold because it captures significantly correlated cases as measured by the SDCC over multiple years. This value was low enough that it allowed for the visualization of significant networks for every year studied yet not too low to relate ever municipality with a dyad. Some years vary drastically in homicide difference from the previous year while others do not, making SDCC quite large for in years with drastic year-over-year changes. A small theta illuminates these important variations. After the truce signed between the gangs in early 2012, there was a huge decline in homicide resulting in large SDCC values that represent highly correlated changes in homicide between paired municipalities. In order to appropriately visualize the reports, I chose a theta of .05 for that year alone 1. Table 3.2 summarizes the total sum of correlated municipalities.

Figure 3.5 is a collection of annual network analyses compared across time to illuminate structural changes in organized crime in El Salvador. Here we can see changes in intensity,

1The change is Theta is only used for ease of visualization, the comparison between years is still made with theta 0.001

55 Figure 3.5. SDCC Network Graphs

56 Table 3.2. Sum of Correlated Municipalities Year Theta Correlation Links 2006 .001 283 2007 .001 114 2008 .001 121 2009 .001 259 2010 .001 62 2011 .001 58 2012* .001 2,676 2012 .05 104 2013 .001 261 2012 Theta .001 excluded from visualization as well as location of homicides which alter the typologies of correlation networks. A con- centration of correlated networks in the central portion of the country is evident starting with 2006. This was expected given that San Salvador, the capital and financial hub of the country, is located in this region. There is a spread of links towards the west side of the country, with only a few connecting with the eastern municipalities. These results are cor- roborated by geospatial analyses. The first step in this process is geo-referencing homicides by municipality, as seen in Figure 3.6.

Homicide counts fluctuate from 0 to 225. The municipalities that have the highest homi- cide rates are located in the center of the country spreading towards the west of El Salvador.

There is one case of high homicide on the north-eastern side of the country but it seems to stand alone. Following the visualization of homicides, I calculate the incremental spatial autocorrelation for each year in order to estimate the scale of analysis. Using the Fixed

Distance Band, I run a hot spot analysis, because it is the best conceptualization of spatial relationships when there is a large variation in polygon size (as seen in the municipality data).

By using the fixed distance with the calculated scale of analysis, each municipality had a neighbor regardless of its size (ArcGIS, 2014). With this information determine existing

57 Figure 3.6. Geo-referenced Homicides 2006 spatial relationships and estimate the Global Morans I to determine spatial autocorrelation per year.

The result of Global Morans I Hot Spot analysis is mapped in Figure 3.8: Cold spots are blue, while hot spots are red. A hot spot is a cluster of municipalities that are grouped together with related homicide counts. Here we can identify statistically significant spatial clusters of high and low spots of homicide. This indicates that the results are not random and are instead clustered around the central region of the country spreading west. Cold spots are regions where municipalities with low homicide counts cluster around each other.

These are mainly located on the north eastern portion of El Salvador. By looking at the

58 Figure 3.7. Global Moran’s I 2006

59 Figure 3.8. Hot Spot Analysis 2006 original homicide map, though, we can tell that homicide is not equally distributed across the country and that some municipalities experiencing high homicide rates are not clustered.

Because the hot spot analysis did not identify a statistically significant hot spot on the northeastern section of the map, this study includes a cluster and outlier analysis. By running

Anselin Local Moran’s I, as seen in Figure 3.9, I identify municipalities in El Salvador that have neighboring features with similarly high or low values (ArcGIS, 2012).

The analysis in Figure 3.9 designates the central region of El Salvador as a cluster of municipalities with high homicide values. Next to this cluster we note an outlier where a

60 Figure 3.9. Cluster and Outlier Analysis 2006 municipality with a low homicide count is surrounded by higher count ones. In contrast, this model identifies the municipality of San Miguel in the northeast section as an outlier with a high-low relationship. This means that this municipality has a high homicide count but the municipalities around it have low homicide counts, helping explain why this municipality is not within a hot spot.

Geo-Spatial analysis of this data helps clarify some of the distinct characteristics of homicide for 2006, but it analyzes the data for only that year. By adding the SDCC network I am able to capture year-over-year differences between municipality dyads. The 2007 network map, shown in Figure 3.10, maps the homicide correlations from the change between 2006 and 2007. This network is made up of 114 links and 42 nodes, making it easy to visualize.

61 Figure 3.10. SDCC for 2007

Degree, closeness, and betweenness are centrality measures used in SNA methods to help us understand patterns in the data. These methods are specific to dynamic node changes, in this case homicide changes per municipality. Degree is the measurement of a nodes activity and is defined by the number of direct links each node has. A high degree node is considered an important point in the network, yet it is not necessarily a good indicator of leadership (Xu et al., 2004). For the purpose of criminal networks, degree is used only as a measure of activeness. Of the 42 nodes on the 2007 network, San Salvador has the highest degree with 39 direct links. This is not surprising given that it has been identified by both hot spots and cluster analysis as a locality of importance. Nodes for San Miguel and Cuscatancingo municipalities both have 20 direct links, making them the second highest degree nodes. This is an important finding because we now identify San Miguel as one of the second most linked nodes in the country, in contrast with its lone-standing identification in the 2006 cluster analysis. By identifying municipalities with the highest degree of activity

62 we can respond to increasing trends of homicide by controlling these locations with the highest level of known degree. Closeness is actually a degree of farness because it measures how far one node is from other nodes. We use it to identify how well an individual node connects to other members of the network. A high degree of closeness indicates that the node is the most central one in the network, and therefore closer to all other nodes (Xu and

Chen, 2005). In the case of El Salvador, San Salvador happens to be the node with the highest measure of closeness. Even though this is simply a function of geographic location, it does affect the connectivity of this node to all other nodes in the network. By identifying

San Salvador as the node with the highest closeness value we can predict that it will also be a node with high homicide activity and can also consider it a transition node for gang responses. The last centrality measure betweenness, which calculates the number of times a node acts like a bridge between other nodes. It is important to note that betweenness is based on the shortest path approach, meaning that it joins nodes that would pass through it to connect with each other the fastest possible way (Xu and Chen, 2005). When dealing with criminal networks this node is considered a catalyst of activity and its neutralization could destabilize the entire network. Unsurprisingly for El Salvador, San Salvador is the node with the highest measure of betweenness. This would indicate that simply because of convenience San Salvador will be a platform for crime distribution. These three measures are an indication of what patterns gangs have been using to respond to violence. It is known that an attack on a specific municipality may require the retaliatory attack on a different, and in many cases unrelated, municipality. By understanding how these networks operate we can get a better understanding of what a potential target could be.

Discussion

There are multiple methodological approaches to analyzing criminal organizations; this paper uses social network analysis to identify municipalities related by the difference in homicide counts. By calculating the single datum correlation coefficient, municipalities can

63 be paired to capture interconnected change in homicides between one year and the next.

Going beyond the identification of sensitive regions, I also identify how some of these clusters may be connected across the country. I complement this approach by using geo-spatial analysis where related municipalities are identified according to location. The approach used in this paper extends geo-spatial analysis by deviating from the idea that criminal activities spread via adjacent municipalities. El Salvador is a small country with a long history of gang activity (Cruz, 2011). These gangs tend to operate as interconnected systems where a central leader, many times located within a prison, coordinates activities for the entire country. This interrelated arrangement highlights the connectivity of criminal violence, such as homicide, beyond geographically adjacent schemes. Given that gangs in El Salvador take orders from prison leaders identifying a pattern of criminal response could be beneficial for state response. The SDCC network is also beneficial for pattern analysis through time because it uses the difference of homicides between years, taking into account the temporal relationship of homicide. The visualization of this network is helpful for evaluating the temporal change in links and the topology of the network itself. By evaluating the network through time we can observe expansion or contraction of homicide links and how gangs are responding to attacks.

The visualization of SDCC through time is very informative because it links municipalities that appear to have independent issues on the surface. This is observed in Figure 3.5 where there are networks focused in the central region of the country in 2006, but expanding to the western part of the country by 2008. Centrality measures from SNA establish San

Salvador as an important hub for this criminal association. The evident shift of municipal nodes towards the eastern portion of the country is associated with government instability in Honduras and new availability of illicit activity paths through this area. Another visual change in network connections is evident from 2011 to 2013 where the network was sparse, then expanded significantly. This is directly related to the period before the truce between

64 the gangs, the truce agreement, and its immediate effect. Homicide rates were very high in 2011 but dropped drastically in 2012 for all municipalities. This synchronized drop in homicide is evident in the data with high measures of SDCC and the highest count of correlated municipalities. Networking models are a new approach to analyzing criminal organizations and events. Even though there is still room for development with these methods, they do introduce new patterns in criminal relationships. This is evident in homicide counts in El Salvador were municipalities that were not otherwise connected are identified as significant nodes, then later identified by more commonly used methods.

65 CHAPTER 4

ESMED

EVENT DATA OF MARAS

Authors – Viveca Pav´onTercero

The Department of Public Policy and Political Economy, GR31

The University of Texas at Dallas

800 West Campbell Road

Richardson, Texas 75080-3021

66 Abstract

The 2012 Peace agreement in El Salvador between Mara Salvatrucha (MS-13) and Barrio

18 was an unprecedented truce between two of the deadliest rival gangs in Latin America.

This agreement serves as an indication of cooperation with a greater purpose between groups

that had been deemed unorganized. This raises concerns of what circumstances will facilitate

cooperation between rival groups and what can be expected of this collaboration. Following

the truce there was a genuine fear that gangs in El Salvador could become political actors by

using Violent Lobbying and/or Violent Corruption. Even though the gangs have not proven

to have political power there is still a fear that they can continue to influence politicians

behind closed doors. This paper seeks to identify the tactics the gangs could use if they did

attempt to lobby the state. These specific violent acts help us understand what pressures

the government of El Salvador could have been facing that led to the truce being agreed

upon. It also distinguishes what changes in violence the country experienced during and

after the agreement. In order to identify these specific patterns I use event data analysis

on newspaper stories from El Diario de Hoy to help identify is any coercive methods where used by las maras.

Introduction

Central America is the deadliest region in the world. The UNODC reports that the level of violence in this region is higher than in any single nation, including those at war. In

2017, Central America had an annual homicide count of 14,575 people, with 13,129 of these casualties in El Salvador, Guatemala, and Honduras the Northern Triangle (EFE, 2018).

This staggering number is far beyond even the average civil war death toll of 10,500 (Lacina,

2006). The countries of the Northern Triangle have experienced war in recent decades and are unfortunately familiar with high homicides rates, yet current homicide figures are alarming.

El Salvador’s 2015 homicide rate of 108.6 people per 100,000, for example, is the highest for any country in nearly 20 years (Bank, 2018). Drug Trafficking Organizations (DTOs) and

67 gangs are responsible for most of the bloodshed in the Northern Triangle (Demombynes,

2011). Although there is some trafficking through El Salvador the country can attribute most of its crime and violence to gangs (Garzn, 2014). This is not surprising given that it has become an epicenter of gang activity (Cruz et al., 2000; De Cesare, 1998; Dudley,

2014; Smutt and Miranda, 1998). It is estimated that there are 100,000 gang members

(Lopez, 2016) in the Northern Triangle with about 60,000 active members in El Salvador alone ((ICG), 2017).

There are two prominent gangs in the Northern Triangle: Mara Salvatrucha (MS-13) and the 18th Street Gang (Barrio 18 ). Even though both gangs are of great interest within

Central America MS-13 has received more international attention. Their global presence has extended continents and is now considered a threat in countries outside of Latin America.

Their recognition and menace is such that US President Donald Trump has launched a hate campaign labeling them as animals of MS-13 (White House, 2018). Unfortunately, there is not much known about these gangs at the transnational level. There are studies on MS-13 but these have focused on their operation in El Salvador (Dudley and Silva Avalos, 2018). I propose a transnational study of Maras via event data with geo-referencing.

This project consists of generating an ontology on gangs, specifically transnational gangs from El Salvador. It is based on the methods used for the creation of OCVED using EVEN-

TUS ID. The following definitions will serve as the foundation for the generated variable of this study. Violence will denote the intentional infliction of physical harm on people or their property for tactical or strategic motivations (Kalyvas, 2006). A Mara is defined by the UNODC as a street gang with territorial control, for this study we will focus on Mara

Salvatrucha 13 and Barrio 18, and its known factions.

Literature Review

Gangs from El Salvador, mainly Mara Salvatrucha, have gained international recognition for their atrocious criminal activities. Within El Salvador they are labeled as terrorists and

68 are considered public enemy number one. The US Treasury Department designated MS-

13 as a transnational criminal organization (Dudley and Pachico, 2013). The current US government considers them a top priority for immigration and security policy. Yet there is still a lot that is unknown about the gang and its hierarchical structure. But there does not appear to be a transnational hierarchy that matches the perception of organized collaboration. Instead every country has its own clica leaders that makes national decisions but can communicate with other country leaders and even international comrades (Martinez et al., 2012).

Gangs in El Salvador have proven influence over society through their violent acts, but they have historically lacked the leadership required for true political power. La Tregua in

2012 challenged this notion though. The truce made it evident that gangs could cooperate, could make arrangements from within the prison system to be followed in the streets, and could have government officials agree to meet some of their demands. This level of organiza- tion and the government’s willingness to cooperate made people nervous about the political gain gangs would earn and the power they could leverage (Dudley and Pachico, 2013).

This research on local gangs will generate original event data to explore the spatial- temporal patterns of conflict in multiple countries. Most accounts of violence in Latin

America rely on national homicide statistics, which do not differentiate between types of crime or who the perpetrator was. While homicide statistics are relatively reliable and readily available, they do not do a good job at distinguish gang-related violence. The case of El Salvador is an example of this issue. Even though the National Civilian Police did start keeping record of homicides attributed to gang in 2004 there is a discrepancy between what is reported in news reports and believed to be crimes attributed to gangs, and what is reported by the police. Figure 4.1 shows homicide and homicide attributed to gangs.

Note how on 2012 there is a significant drop in homicide but homicides attributed to gangs stay constant. This is interesting given that the truce between the gangs was in effect that

69 Figure 4.1. Homicide and Homicide by Gangs Source: (INCSR, 2014) year and most scholars and area specialists agree that this decline in homicide was due to a decrease of crimes committed by gang members. Overall not many homicides are attributed to gangs according to police reports, but this does not seem to be what people experience given the huge impact of the drop in homicide once the truce began.

In El Salvador the police has been a big target in recent years. Figure 4.2 shows this increase in police attacks following the dissolution of the truce. There is a real fear in communities were police officers live that mareros will attack the entire neighborhood in order to seek revenge from the police. Yet national data available does not indicate that police officers are continuously being attacked. Instead we can see a huge spike in 2015 but then a much lower rate for the years to follow.

This lack of reliable and consistent information prevents local authorities, policy experts, and the international community from fully understanding the problem and implementing

70 Figure 4.2. Police Officers Murdered Source: (Diario1, 2016; Marroquin, 2017; Focus-Data, 2017)

regionally specific policies to combat the gang issue. One main goal of this study is to create

event data that better describes maras and the nature of their criminal behavior. Because measures of this nature have not been readily available researchers have used less than perfect proxies. One measure used, for example, is homicide counts. In this case researchers use trends are guidelines and not the raw figures themselves. Recent studies within the region have disaggregated intentional homicide and found some interesting trends. One trend is the age and gender distribution of homicide. The 2013 Global Study on Homicide indicates that

79 percent of all global homicide victims are male and that approximately 73 percent of the global homicide victims are between the ages of 15 and 44. For Central America and South

America, the trends are magnified with male victims between the ages of 15 and 29 having a homicide rate four times greater than the global average (UNODC, 2013). The same results were reported in the 2011 Crime and Violence in Central America report presented by the

71 World Bank indicating that 60 percent of all homicide victims in the region are men between the ages of 15 and 34. These findings have been used as a proxy to identify gang or drug trafficking related violence as any homicide that occurs for males between the ages of 15 and

34.

Another interesting trend of homicides in the Latin America is the mechanism used. It is estimated that 30 percent of homicides in the Americas are attributed to organized crime and/or gangs (UNODC, 2013). With 66 percent of the homicides perpetrated by firearm, the Americas have the largest prevalence of firearm killings in the world. Even though it is inconclusive as to why the majority of the killings are by firearm, authors such as Cuevas and Demombynes identify the availability of firearms from illicit markets as a big driver for this type of crime (Cuevas and Demombynes, 2010). It also happens to be the preferred method of criminal operation by many organized groups (Breve-Travieso, 2007). Scholars have used this information to disaggregate homicide even further and attribute the death of males, between the ages of 15 to 34, with the use of a firearm as gang related crime.

A big problem with the previously mentioned qualifications is the inclusion of false pos- itives. Unfortunately, these measures do not uncover the true nature of a murder, they are simply a matter of age, gender, and weapon used. This generalization can capture events which are not directly related to either gang or organized crime. The use of this proxy is op- portunistic and irresponsible, yet many non-governmental institutions rely on this measure.

The use of event data can help us gain a better understanding of gang attacks and their use of violence. This can help us answer questions like: Do gangs attack without logic? Or are they to disorganized to successfully influence the state? DO gangs engage in organized crimes to gain political power? Do they target state officials? It is undeniable that gangs use violence to obtain territorial control and to maintain a form of power over society, but I want to know to what extent can they actually influence state responses.

Method

72 I propose the use of supervised machine coding in Spanish to ‘read’ newspaper stories and identify criminal events with gang characteristics. This study consists of generating an ontology on gangs, specifically transnational gangs from the Northern Triangle. It is based on the methods used for the creation of OCVED, Organized Criminal Violence Event Data in Mexico 2000 – 2010, using EVENTUS ID a software program to machine code conflict events from Spanish text (Osorio, 2013), specifically gang and drug violence events in Spanish language media and press releases (Schrodt, 2010; Osorio, 2013). This project would expand on the OCVED model in two ways: first it adds to the original ontology by including gang related acts and actors; second, it expands its geographic analysis by evaluating El Salvador.

One of the main characteristics of the OCVED database is that it distinguishes itself from government generated reports by not just counting DTO related Homicide but by also being able to identify who did what to whom, when and where. With this distinction event data sets itself apart and improves the quality of the information available to researchers by adding specific characteristics of events to our quantifiable knowledge. OCVED does not just report homicide, it also reports events related to organized criminal violence such as extortion, kidnappings, and money laundering. This database is also able to identify trends in modus operandi. While a homicide report indicates a count OCVED can give counts for specific deaths, such as be-headings or hangings. This type of information provides details on the modus operandi of a criminal group and its evolution. These characteristics will become relevant in the case of El Salvador where we can evaluate gang specific related events events.

After several meetings with the archival department of El Diario de Hoy in El Salvador

I was granted access to their files. I also downloaded and converted 99 stories be used as a manageable data set to start training Eventus ID for larger corpus of text. An important component of the software is that it requires rigorous work on verb and actor dictionaries as well as location dictionaries specific to each country. Figure 4.3 illustrates the event coding process using this software.

73 Figure 4.3. Eventus ID Coding Process Source: (Osorio and Reyes, 2016)

As we can see in Figure 4.3 this event data coding protocol uses six stages: 1) gather information from newspapers following the desired inclusion and exclusion criteria; 2) trans- forming it into an Eventus ID readable file; 3) run Eventus ID to identify events after the creation of actor and verb dictionaries; 4) geo-reference the dataset using the location of the reports generated in stage three through the use of a location dictionary; 5) validate the data by comparing a sample to a section manually coded by humans, recode the dictionar- ies if needed, run this process over until there is an accepted level of convergence between machine and human coding; finally, 6) generate a dataset of validated geo-referenced event

74 data (Osorio and Reyes, 2016). This type of event identification is a better method for

finding gang related events because it only includes crimes that follow gang modus operandi.

Through this methodology gang related violence can be given their own criminal categories

(e.g., extortion, murder, intimidation, etc.).

EVENTUS ID - Gangs in El Salvador

Stage 1.

Stage 1 of Eventus is about information gathering. As previously mentioned I collected stories from El Diario de Hoy. I have access to over 14,000 newspaper pages from 2005 to

2016, but the original query for selecting these stories included organized criminal activity that is beyond the scope of this study. I ran a new query focusing specifically on Maras and used a sample of 1,000 stories to test the conversion. Unfortunately, El Diario de Hoy stores its daily newspaper in a full sheet PDF, they basically captures an image of each newspaper page and store it in PDF formant. This means that every PDF I have access to has two to three different stories within it with different fonts, images, and test orientation. The basic features of a newspaper page make it difficult for automated extraction of individual stories given that the software would need to identify were a story begins and were it ends even if this is on a separate column on a different side of the newspaper. Figure 4.4 is one of the newspaper sheets being used for this study, it has three separate stories and three images.

A difficult task for automated conversions from PDF is identifying the sequence of the text which starts from left to right but with a division of stories identified by a heading.

Using Adobe Acrobat DC Pro I was able to convert each of these newspaper sheets from pdf to html. I then separated each identified story originating from its heading and text orientation into an individual reported story. After this conversion I was able to format each story for EVENTUS ID to read. A critical component on stage 1 is the naming of each story. The nomenclature consists of the date, the acronym used for the source and the counter for unique identification of news stories. The date uses the format YYYYMMDD,

75 14 NACIONAL SUCESOS EL DIARIO DE HOY Jueves 7 de abril de 2011 [email protected] En kilómetro 13.5 del Anillo Periférico Unmuertoyunlesionado enataqueahermanos » En Juayúa matan FOTO EDH / MAURICIO CÁCERES a un estudiante LOSTRESSUJETOSserán procesados en los tribunales por el delito de receptación o adquirir objetos de dudosa procedencia, según la Policía. de bachillerato Jaime López/ Emilia Pacheco Víctimadescubre Tony Francisco Chanta de 28 años y su hermano de 20 fueron atacados por varios sujetos arma- do, ayer por la mañana cuando se suautorobadoen dirigían a su trabajo en el kilóme- tro 13 y medio de la carretera Ani- llo Periférico en Ciudad Delgado. En el incidente; el primero tallerclandestino murió y el segundo resultó grave- mente lesionado quien fue trasla-

dado a un hospital de San Salva- FOTO EDH / MAURICIO CÁCERES establecimiento en busca de un dor donde se recupera de sus he- TÉCNICOS de laboratorio técnico científico de la PNC recogen las evidencias del » Roban cuatro servicio automotriz y se encon- ridas, según informó la Policía de asesinato de Tony Chanta, registrado en el kilómetro 13 y medio del Anillo Periférico. vehículos al día tró con piezas del carro que dí- Ciudad Delgado. as antes le fue robado. Poco después familiares que en San Salvador De inmediato la víctima avi- llegaron a la escena del delito, no rrió a 50 metros de la caseta poli- só al 911, y luego un equipo de in- se explicaban por qué fueron ata- MATAN POR NEGARSE cial. En la escena se visualizaban al vestigadores de la Delegación cados los muchachos, excepto A LAS PANDILLAS menoscincocasquillosdearmade Jaime López de Mejicanos comenzaron la in- que la noche anterior un grupo de fuego, calibre desconocido. dagación que terminó con la lo- Algunos casos son; un sujetos se acercó a discutir con En otro caso de homicidio, un Tres sujetos fueron capturados calización del taller, la captura estudiante del Instituto uno de los jóvenes. Nacional de San Bartolo, taxista fue masacrado a balazos en el interior de un taller clan- de los sujetos y el decomiso de No descartan que este grupo Ilopango, dos señoritas en presuntamente por un sujeto que destino dedicado, según agen- dos vehículos robados entre el 11 un día después les haya seguido escuela La Unión de minutos después fue detenido tes del Sistema 911, a desmante- y 21 de marzo en los municipios los pasos hasta asesinar a uno de Tonacatepeque y un joven en porlaPolicíaNacionalCivil,enla lar vehículos robados en la peri- deSanSalvadorySanMarcos. ellos. Las víctimas quienes traba- Lirios del Norte, Calle 5 de Noviembre. El taxista feria capitalina. Los detenidos El inspector Juan Bautista jaban en construcción, residían Cuscatancingo. Lazaro Granado, de 58 años, fue son Pedro Corea, Arturo Ama- Rodríguez, jefe del 911, manifes- en la colonia Valle Verde de Apo- asesinado frente a una gasolinera. ya y Carlos Muñoz. tó que los detenidos son parte pa y diariamente salían a su traba- En otro hecho criminal, Carlos El taller fue localizado en fi- de una supuesta estructura de- joabordandolaRuta38. Eduardo García de 16 años, quien nal 13a. Avenida Norte y Calle dicada al robo y hurto de vehí- Acostumbraban bajarse en el culo, para luego dirigirse al lugar estudiaba primer año de bachille- San Carlos, tres cuadras al po- culos con miras a desmantelar- paso a desnivel situado en el kiló- de trabajo, cerca del monumento rato en el Instituto Politécnico de niente del Cuartel San Carlos en los en piezas, para luego vender- metro 13 de la Troncal del Norte de El Salvador del Mundo. En otro Sonsonate fue asesinado en Juayúa San Salvador. los como repuestos. para luego caminar hacia la pasa- hecho ayer a las 8:30 de la mañana el martes a las 7:00 de la noche. Según el 911, el referido taller Rodríguez afirmó que hay al rela de Ciudad Futura II, en Cus- un panadero, de 25 años, fue acri- Según las primeras pesqui- fue encontrado por una persona menos siete sujetos de esta ban- catancingo. En ese lugar se encon- billado en el kilómetro 59 de la ca- sas policiales, el joven fue ulti- a quien sujetos armados en fe- da por capturar. A diario el 911 traban con su patrón quien siem- rretera que de San Salvador con- mado porque se negó a pertene- brero lo despojaron de su vehí- registra el robo de cuatro vehí- pre pasaba a recogerlos en su vehí- duce a Santa Ana. El crimen ocu- cer a las pandillas. culo. Al parecer la víctima llegó culos en la capital.

Muere ex locutor radial en accidente El Diario de Hoy control del auto placas P-478-57, se firmará o desvirtuará con la autop- salió de la calle y chocó contra tres sia si Aparicio Hernández, había Un ex locutor radial murió ayer al árboles que se encuentran sobre la ingerido bebidas alcohólicas. accidentarse el vehículo que mane- acera. Al chocar con el último ár- Además, los policías de Tránsi- jaba, aparentemente por conducir bol, el vehículo volcó a raíz del to establecieron que el exceso de a excesiva velocidad, según infor- fuerte impacto, por lo que hubo ne- velocidad y la distracción del con- mó la unidad de comunicaciones cesidaddequeelCuerpodeBom- ductor habrían sido las causas del de la Policía. beros utilizara equipo especial pa- accidente. Durante el reconoci- El accidente se registró a las 5:10 ra rescatar el cadáver de Aparicio. miento de la víctima, los policías le de la mañana sobre la 25a. Avenida El vocero policial explicó que hallaron una memoria USB que la Sur a inmediaciones del cemente- los agentes de Tránsito que hicie- tenía en su mano derecha, lo que rio La Bermeja. ron la inspección encontraron va- hace suponer a los policías que se FOTO EDH / MAURICIO CÁCERES El informe policial señala que rias botellas para cervezas vacías descuidó cuando trataba de conec- BOMBEROS TRABAJAN en la recuperación del cadáver de Emilio Ernesto Emilio Ernesto Aparicio Hernán- dentro del vehículo. Sin embargo, tarlo a un aparato para escuchar Aparicio Hernández. Quien murió en un accidente vial registrado en 25a. Av. Sur. dez, de 29 años, habría perdido el el Instituto de Medicina Legal con- música.

Figure 4.4. El Diario de Hoy newspaper sheet sample with YYYY indicating the four digit year, MM is the two digit month, and DD is the two digit date. Following the date is a three digit counter, ccc. Lastly the source is identified by an acronym. yyyymmddccc SRC.ext The nomenclature clearly identifies each individual story and is a necessary component for Eventus ID to deal with duplicates later on.

Stage 2.

In order for Eventus ID to more accurately identify events the input files should be divided by paragraphs. The belief is that smaller coding units reduce error, and for this reason each

76 Figure 4.5. web2eventus Output news story is fed through a software that divides it by paragraph. Web2Eventus take each properly labeled .html file and splits it by paragraph and assigns a unique identifier to each paragraph. It also assigns a global identifier in order to identify the paragraph location for all of the paragraphs identified before. These identifiers proceed the nomenclature of Stage

1. Figure 4.5 is an example of web2eventus output.

In the indicated line we can notice the date (20150815), followed by the nomenclature

(20150815001 SAL). The addition of the paragraph number is next with the labels for the in-article paragraph ( P5) count and the global paragraph count ( P52).

Stage 3.

Stage 3 consists of event coding. The software uses pattern recognition to extrapolate from the text corpus. In order to do this a list of proper nouns and verbs needs to be added with the corpus for identification of the words or phrases. With this process the software is trying to identify who did what to whom (where and when are identified in later stages).

Eventus ID looks for these patterns within the same paragraph for it to count as an event, it

77 assumes that the event will be stated in one paragraph. The word sources for identifications are called Dictionaries and are further explained below. The first component of an event is normally an actor doing an act to somebody or something. Event data calls these initiators the Source. The Target is the person or object upon whom the source is carrying an action. Due to the nature of violence both source and target are coded in the same dictionary. Table 4.1 is a list of the categories that this study uses.

Table 4.1. Actor Dictionary Categories Category Code Army 10100/10200 Asset 80100 - 80600 Communications 90100 Criminals 40100/40101 Drugs 50100 - 50600 Person 30101 - 30301 Police 20100 Vehicle 70100 - 70700 Weapon 60100 - 60700 Other 100100/100200 Source: adapted from (Osorio, 2013)

The dictionary used consists of over 640 actors. The addition of code 40101 under criminals is the inclusion of maras into the dictionary. This serves as the introduction of the gang typology into OCVED changing it to El Salvador’s Mara Event Data (ESMED). This dictionary also added upon the existing one by including specific actor subcategories such as students, teachers, and cooks. This became necessary due to the descriptive language that used in the media in El Salvador and how they portray an event. Another addition was the inclusion of children as a subcategory. This is also specific to this country, were crime has escalated to the extent of including children and police forces as both targets and sources. Each actor in the dictionary is given a numeric code that corresponds to the categories presented. Figure 4.6 is an example of actor coding.

78 Figure 4.6. Example of Actor Coding

79 One of the main components of event data is identifying the action taking place. A verb dictionary is used to identify these actions. In this research over 500 verb phrases are used, most of them corresponding to violent actions. Table 4.2 is the list of categories used.

Table 4.2. Verb Categories Category Code Wound 100 Steal 200 Shoot 300 Kill 400 Arrest/Seize 500 Burn 600 Source: adapted from (Osorio, 2013)

Due to the complexity of the Spanish language a majority of the verbs are accompanied by a set of words that help refine the meaning of the verb according to the context, they make up what we address as verb phrases. Each word is connected by an underscore ( ) to indicate a sequence of words that should be read together in order to be identified by Eventus ID. There is also an asterisk (*) used to indicate the position were a verb can be input in order to create a set of word combination. Just like for the actor dictionary each verb is coded with a set of numbers that corresponds to the category code. An example of these can be seen in Figure 4.7. The verb dictionary used for the case of El Salvador includes words and phrases specific to the country and to the language used to describe violent events. These verb patterns are an important aspect of ESMED because they contain verb phrases specific to gang activity adding to this new typology. Stage 4. Stage 4 focuses on event location. Through the previous stages Eventus ID is capable of identifying the source, action, and target of an event but identifying location is more complex. The program uses the identified events as the starting point (stage3) then proceeds to load

80 Figure 4.7. Example of Verb Coding the location dictionaries to identify location within the same paragraph that the source, action, and target were found. If the program does not find a location within the same paragraph it proceeds to look for it in the entire document.

Location Dictionaries

ESMED geo-referenced the data at the municipality level. This means that two separate dictionaries are used, one that recognizes the State, and a secondary one to identify the

Municipality. Each state is numerically coded, and each municipality is uniquely coded by state number first followed by a three-digit counter. Figure 4.8 is an example of the state and municipality coding were the State of Cabaas is coded as 4 and its municipalities have the numbers 4001 to 4009. The location dictionaries are supplemented by a filter dictionary were places and things are identified as separate.

81 Figure 4.8. Example of Location Dictionary

Once Eventus ID has geo-coded the available events it generates a coded output. This is raw event data, as seen in Figure 4.9.

This output is further processed in STATA, where codes can be replaced by subcategory labels and these can be aggregated to main categories. The date is disaggregated to be able to identify when an event occurred. An issue that is addressed in STATA is the matching of state and municipality. Because the Eventus ID can search for a matching municipality within the entire document there are seven possible municipality identifications that can take place but it should only include those were the initial municipality code number matches the state code. Given that the corpus is input at the paragraph level there is the possibility of duplicates within the stories, were the same coverage is repeated in the article or multiple articles are about the same event. This issue is prominent in event data and Eventus ID deals with it by removing the same event on the same date occurring in the same municipality.

The clean-up process also includes dropping instances where there is no event, meaning the software did not identify a source, action, and target within a paragraph. The intended output is one were an event is clearly identified with geo-location. Figure 4.10 is the output of

82 Figure 4.9. Example of Raw Output

Figure 4.10. STATA Geocoded Output geo-coded events after code transformation. The actor in this case have not been aggregated to their main categories.

83 EVENTUS-ID Counts The event output is analyzed further to be able to identify who the perpetrators are, what is their main action, and who are the targets.It is trying to answer: Who committed the event? (Source) What was the event? (Action) Onto Whom? (Target) Where did the event happen? (Location) As previously mentioned I collected stories from El Diario de Hoy in PDF format and converted a sample of these to HTML. The results for this transformation are faulty and are not a good representation of event data gathering using EVENTUS ID 1. In order to properly test the software and make sure the new ontology is working I ran a query on the newspaper website to extract the stories directly and transform them into html with the WebTextDownloader software included with Eventus ID (Osorio, 2013). Using the word Violence (violencia) in El Diario de Hoy’s search engine I extracted the first 100 stories that had a title related to gang and/or violence. This process helped with the testing ESMED but also highlighted a need for better transformation into html. It is a basic process that will need to be improved for the creation of the larger corpus but it works well for data training given human discrimination capabilities. Of the 100 stories that were used Eventus ID identified 20 fully geocoded events at the municipal level. Because there are several reports within each newspaper report they tend to group them by location and do not explicitly mention this location for every event. In order to get a larger sample of events I geocoded at the state level since these are mentioned at a greater rate within the reports. With this change Eventus ID identified 88 events. The dictionaries were aggregated to a few categories for a better visualization and count of what and who the events were about. “Person” is anyone that is identified yet not categorized as a government official, police, or criminal. Figures 4.11 is the frequency of the main source in the events. The software identifies “persons” as the main actor in most

1The results from this test are available in Appendix C

84 Figure 4.11. Main Source of the Events events. The actions were also aggregated. Figure 4.12 is the frequency of the main action taking place in these events, with killings being the main activity. Figure 4.13 is the main recipient of the actions, here we see that people are the main target while weapons seems to be identified as targets of events as well. Figure 4.14 is geolocation with the exclusion of municipalities, here we note that Eventus ID identified San Salvador as the state with the highest occurrence of events, which is expected given that it is where the capital is located and that it is a state with high population density.

This analysis highlighted some concerns with the output. First, if we observe the list of events we can see that several happened on the same date. Because the corpus is analyzed by paragraph Eventus ID is identifying multiple events within the same news story. The 88 events identified were all out of 25 articles. This means that even though 100 newspaper stories were input only 25 of them had enough information to identify a fully geocoded and

Source-Action-Target event. Some news stories lack information and I believe there is a large degree of separation between were a municipality is mentioned in the article and were the

85 Figure 4.12. Main Action of the Events

Figure 4.13. Main Target of the Events

86 Figure 4.14. Main Location of the Events - State Level state is mentioned. Even though this modification proved to be a better fit regarding location there are still some improvements that could be made regarding location identification.

Validation

The validation for this paper is two-fold, external and internal. Externally OCVED has been compared to other event data sources, and it has an accuracy of about 71 percent. This is around what is expected for computer coded event data (Schrodt, 2010).

For internal validation I coded the events that Eventus ID recognized as fully geocoded.

Figure 4.15 shows the comparison at the municipal level. Of the 20 events identified, some within paragraphs not by unique news story, only 1 was a full match between human and machine coding. This clearly indicates that there is an issue with the validity of the event output. Some events would have been improved if the municipality was not included, but for validity purposes it was included as a direct comparison on what OCVED codes versus the gold standard of human coding.

ESMED Concerns

87 Figure 4.15. Validation - GSR v EVENTUS ID

Despite OCVED being a good measure of violent events in Mexico it seems to fall short for the case of El Salvador and gang activity. There are several reasons why it does not seem to fit:

1. Sources. OCVED uses a variety of sources for its analysis. The database relies heavily

on police sources, which are not readily available for El Salvador. This source works

well for OCVED because it is written in a more direct format. Newspapers in El

Salvador appeal to sensationalism and storytelling. Instead of direct reporting they

try to engage the reader and have them sympathize with the victim (target in many

cases). This becomes a big limitation for ESMED because it creates an even larger

complexity of the language.

2. Language. As stated in the previous limitation the use of complex language makes it

difficult for Eventus ID to classify sources and actions. Many articles in El Salvador do

not describe an actor as simply a person, or a member, or a police officer; instead, the

description is drawn-out. For example a man would not just be described as a man, or a

teacher (his occupation) but he would be described as “a fifth grade teacher for school,

a father of three. This string of words is too descriptive to be included in the dictionary

88 which makes it difficult to identify. Eventus ID might classify teacher, school, or father

as the target in this case and these could be classified as either property or a person.

3. Multiple Stories. With El Salvador being the deadliest country (by homicide) in the

world there are days were not every event gets its own story. Instead there is a growing

trend in newspaper reporting to include many events within the same story. Article

20160128001 SAL for example, has 7 homicides reports within it, these are grouped

by similarity in either modus operandi or location. Eventus ID is identifying them by

paragraph but it is designed to collapse these events by source, action, and target rep-

etition or by location, which means many individual events are counted as duplicates.

4. Municipality. A story can have multiple locations included given that many events are

reported in the same article. The software is currently reading the same paragraph

for second order location but if it does not find one it searches for a match within the

entire document. This has made it difficult for Eventus ID to recognize which state and

municipality combination to attribute to each event. In many of the events identified

the first location order (state) was accurate but the second location was not. This

forced the program to drop the event. This was improved when municipalities were

excluded from geocoding but there were still issues with the location. Some articles

report the title story then state “in the same location or “in its vicinity to account for

a new event. Eventus ID cannot find the location for these.

Figure 4.16 article samples are a comparison of El Salvador news reporting on the left and Police reporting for Mexico used by OCVED on the right. The clarity of the language for the police report used for OCVED is evident, as well as the direct description of the event. The article for El Salvador has its issues and differences in reporting pointed out.

Discussion

89 Figure 4.16. ESMED v OCVED News Sources

Preliminary event data result show that individual citizens not necessarily associated with gangs are the ones committing the majority of the crimes in the stories used for this research. This, along with the national data on homicide Figure 4.15 could be an indication that even though the gangs did initially act as violent criminals and have engaged in forms of violent lobbying they do not operate as organized criminal actors with the capability of gaining political power. They reached a truce with each other and the government due to third party involvement but have not been able to reach any other type of agreement on their own. Even though there could be indications of violent reactions, like the increase in police killings in 2015 shown in Figure 4.2, these have not coercive enough to grant gangs lobbying power in El Salvador.

90 Figure 4.17. Homicide Distribution

91 CHAPTER 5

CONCLUSION

El Salvador is a country with a recent history of civil war that despite having reached peace is still experiencing high levels of violence. One of the biggest issues the country faces is violence attributed to gangs, or Maras. The two main gangs of the country Mara Salvatrucha 13 and Barrio 18 migrated into the country in the late 1990s from Los Angeles and have transformed into prison based criminal organizations. Their level of sophistication is quite low yet the level of violence and fear they can instill is very high. These gangs originated in Los Angeles but their change from street gangs to criminal groups occurred within Central America. This research introduces policies that helped transform the maras in El Salvador. It also presents a case on how the gangs do not follow traditional criminal patterns such as nearest neighbor and are instead operating on a reactionary model associated with crimes that occurs in other municipalities. I also present a study on how much power can be attributed to these groups, following the truce between the gangs and the state the gangs changed their modus operandi, was this an attempt to gain power or was it retaliation? The first section of this research looks at the timing of policy implementation in El Salvador. Using break point analysis I highlight how repressive policies in the country happen to be applied shortly before election periods. Unfortunately, El Salvador has been a country where the government has not been able to maintain a monopoly of power and instead have used the perception of fear to highlight how criminal groups may be the true powerful agents regarding violence. This fear is used along with policy implementation to maintain political control. The right wing party ARENA used this tactic in the early 2000s when it first launched Mano Dura policies to help combat gang violence. In 2004, an election year, for the first time since the civil war FMLN had a real chance of being a contender for the presidency. ARENA, with the help of the media, launched a campaign against the gangs. This attack came in the form of repressive policies that society backed once they

92 were presented with scenarios of gang violence and atrocious criminal activities. FMLN was quick to oppose mano dura, making it easy for the media to portray them as ’soft’ on crime. Using homicide data, as a proxy for overall violence, in break point analysis we see that the claim of an increase in crime due to gangs was not actually true. Break point analysis outlines identifies 2004, a year after the implementation of punitive policies, as the year when there was a statistically significant change in homicide trends but the change was in the opposite direction then was intended. The plan that ARENA formulated to attack the gangs did serve the purpose of rallying people behind their policies and re-electing them but it did not help solve the true issues in society. Instead, these policies have been identified as causal factors for an increase in crime and the development of gangs into true criminal organizations as we know them today. This research add to the current literature by identifying how the left wing party, FMLN, that originally opposed mano dura ended up using it for political gain just as ARENA had. FMLN came into power in 2009 with the expectation that policies would shift to a rehabilitation approach, unfortunately this was unpopular and mano dura type approaches were introduced again. After the 2012 truce between the gangs, the people of El Salvador did not support any kind of negotiation with las maras. Given a new election period FMLN reverted back to punitive policies in order to gain political favor. Break point analysis identified 2014 as another statistically significant year in homicide data trend change, again in the opposite direction then was intended. When FMLN made it public that they would not negotiate with gangs and would instead use mano dura tactics homicides began to increase. The implementation of repressive policies was favored by the media and yet again society accepted government’s use of punitive actions against the gangs. As political parties in El Salvador fought over power they manipulated the media and influenced people’s perceptions of violence to fit an agenda that promoted ”punitive populism”. Instilling fear in society has been a tactic used by both parties in order to gain and maintain power despite the obvious unintended consequences of these actions.

93 Continuing the analysis of the dynamics of violence in El Salvador I use social network

analysis to identify the temporal relationship between paired municipalities according to

homicide. This method looks at the interconnectivity of homicide counts from one year and

the next between municipalities, meaning that it is evaluating how two seemingly distinct

regions can be responding to each other when it comes to homicide rates. When study-

ing criminal organizations criminal behavior theory tends to follow the nearest neighbor

principle, which states that increase in criminal activity tend to concentrate around a high

criminal unit, in this case a municipality. But this study takes into consideration the hi-

erarchical dynamic of gangs in El Salvador and allows for the analysis of crimes across the

country and not focused on adjacent municipalities. By removing the adjacency limitation

I am allowing homicide data to dictate which municipalities can be paired with regards to

criminal behavior as a response to each other. The idea is to identify which municipalities

are responding to crime in other municipalities around the country. This becomes a critical

aspect of violence in the country because gangs can react to a specific event by attacking

regions where rival gangs operate not necessarily adjacent areas. For example, an attack

against MS-13 may happen in San Salvador but MS-13 can respond by attacking Barrio 18

in San Miguel, a municipality far east of San Salvador. In El Salvador these dynamics are

quite possible given that the country is very small and is very urbanized. Without much

rural areas every municipality can be the recipient of an attack by rival gangs.

Social network analysis is a good tool for understanding gang behavior because it allows

for the interconnectivity of gang leadership and orders. Maras in El Salvador are prison based organizations and in most instances orders are handed down from within the prisons.

This means that an order from MS-13 leadership can originate in San Salvador but be exe- cuted in Ilopango. The response from Barrio 18 can happen anywhere in the country where gang leadership dictates. In this study I conducted a single datum correlation coefficient

(SDCC) to create this network connection. Figure 3.5 has the year specific networks for

94 homicide. This figure is indicative of changes in homicide patterns on an annual basis. The networks created here highlight which municipalities are relevant during a specific year. This is an important component of gang behavior given that it identifies which regions become significant for the gangs by showing where they decide to respond to attacks. For example there are networks grouped in the central region of the country in 2006, but they expand to the western part of the country by 2008. Centrality measures from SNA establish San

Salvador as an important hub for these criminal acts for both years. The evident shift of municipal nodes towards the eastern portion of the country can be associated with govern- ment instability in Honduras and new availability of illicit activity paths through this area.

When gang leaders determine which municipality to attack they pick ones were they can inflict the most damage. In 2008 this was in an area were there was money to be made by extorting those groups using eastern El Salvador for illicit trafficking of goods. Using SNA we can better identify what regions of the country are of interest during a specific point in time and can help policy makers establish areas needing additional backing.

Gangs in El Salvador have challenged state powers with their use of violence, but they have historically lacked the leadership to obtain true political power. La Tregua in 2012 challenged this notion though. The truce served as a platform to give gangs a voice as a unit and it showed that there was some level cooperation that could be achieved between them.

This made society rethink how gangs operated and consider that now these organizations were proving that they could make arrangements from within the prison system to be followed in the streets, and could have government officials agree to meet some of their demands. This level of cooperation from the gangs, and the government’s willingness to coordinate and meet their requests made society nervous about the political gain gangs would earn and the power they could leverage (Dudley and Pachico, 2013). This research used event data analysis to help uncover the tactics used by las maras in El Salvador. This method can be used to answer questions regarding gang activities and their potential according to their modus

95 operandi. Preliminary results on text analysis of newspaper reports from El Diario de Hoy shows that the majority of the attacks seems to be focused on killings of persons and not attacks against the state or state representatives such as government officials or the police. ESMED can be used to better understand gang activities in El Salvador.

96 APPENDIX A

BIC & RSS

Bayesian Information Criterion and Residual Sum of Squares

Figure A.1. BIC & RSS (2002-2017)

97 APPENDIX B

NEWSPAPER QUERY

El Diario de Hoy query and ESMED query

Figure B.1. Newspaper Queries

98 APPENDIX C

EVENT DATA ON CONVERTED HTML FILES

Event data results for 1,100 stories converted from PDF to HTML.

Figure C.1. Main Source of the Events

99 Figure C.2. Main Action of the Events

Figure C.3. Main Target of the Events

100 Figure C.4. Main Location of the Events - State Level

101 REFERENCES

Aguilar, J. and M. Carranza (2008). Las maras y pandillas como actores ilegales de la regin. Technical report, Informe Estado de la regin en desarrollo humano sostenible 2008. San Salvador: IUDOP.

Aguilar, J. and L. Miranda (2006). Entre la articulacin y la competencia: las respuestas de la sociedad civil organizada a las pandillas en el salvador. In Maras y pandillas en Cen- troamerica. Las respuestas de la sociedad civil organizada. San Salvador: UCA Editores.

Albadejo, A. (2016). No life here: Internal displacement in el salvador. Washington, DC..

Alvarado, R. (2011). Geostatistical spatio-time model of crime in el salvador: structural and predictive analysis. Revista de matemtica: Teora y Aplicaciones 18, 325–342.

Amaya, L. E. (2011). Los sistemas de poder, violencia e identidad al interior de la mara sal- vatrucha 13: Una aproximacin desde el sistema penitenciario. Anuario de Investigaciones 2011 , 103–203.

Amaya, L. E. and J. J. Martinez (2015). Escisin al interior de la pandilla barrio 18 en el salvador: Una mirada antropolgica. Polica y Seguridad Pblic (7).

ArcGIS (2012). Cluster and outlier analysis (anselin local moran’s i) (spatial statistics).

ArcGIS (2014). Hot spot analysis (getis-ord gi*) (spatial statistics).

Bagley, B. (2013). The evolution of drug trafficking and organized crime in latin america. Sociologia, Problemas e Praticas 71, 99–123.

Bai, J. and P. Perron (1998). Estimating and testing linear models with multiple structural changes. Econometrica, 47–78.

Bank, T. W. (2018). Intentional homicides (per 100,000 people). Technical report, The World Bank.

Baylora, E. (1983). El Salvador in Transition. Chapel Hill: University of North Carolina Press.

Bejar, R. G. (1998). El salvador de posguerra: formas de violencia en la transicin. In Violencia en una Sociedad en Transicin. San Salvador: Progama de las Naciones Unidas para el Desarrollo El Salvador.

Breve-Travieso, F. (2007). The maras: A menace to the americas. Military Review, 88–95.

Buergenthal, T. (2004). United nations truth commission for el salvador, the. Vand. J. Transnat’l L 27, 497.

102 Carcach, C. (2008). El salvador. mapa de violencia y su referencia historica. informe de investigacion. Centro de Monitoreo y Evaluacion de la Violencia desde la Perspectiva Ciudadana. San Salvador.

Carcach, C. (2015). A spatio-temporal model of homicide in el salvador. Crime Science.

Cawley, M. (2013). The mara women: Gender roles in centam street gangs. InSight Crime.

Cruz, J. M. (2004). Pandillas y capital social en centroamerica. Maras y pandillas en Centroamerica. Padillas y capital social II, 277–326.

Cruz, J. M. (2005). Los factores associados a las pandillas juveniles en centro america. Estudios Centroamericanos (685-686), 1155–1182.

Cruz, J. M. (2010). Democratization under assault: Criminal violence in post-transition central america. Vanderbilt University.

Cruz, J. M. (2011). Criminal violence and democratization in central america: The survival of the violent state. Latin American Politics and Society.

Cruz, J. M. (2012). The political workings of the funes administrations gang truce in el salvador. Technical report, Improving Citizen Security in Central America: Options for Responding to Youth Violence, organized by the Woodrow Wilson International Center for Scholars.

Cruz, J. M. and M. Carranza (2006). Pandillas y polticas pblicas. el caso de el salvador. In Juventudes, violencia y exclusin. Desafos para las polticas pblicas. Guatemala: MagnaTerra Editores, S.A.

Cruz, J. M. and N. Portillo Pea (1998). Solidaridad y violencia en las pandillas del gran San Salvador. Ms all de la vida loca. San Salvador: UCA Editores.

Cruz, J. M., J. Rosen, L. E. Amaya, and Y. Vorobyeva (2017). The new face of street gangs: The gang phenomenon in el salvador. Technical report, Florida International University and Fundacin Nacional para el Desarrollo (FUNDE).

Cruz, J. M., A. Trigueros, and F. Gonzales (2000). El crimen violento en el salvador. factores sociales y econmicos asociados. San Salvador: Banco Mundial y Universidad Centroamericana.

Cuevas, F. P. and G. Demombynes (2010). Crime and violence in central america. The World Bank: The Central America Unit Poverty Reduction and Economic Management Unit (56781).

De Cesare, D. (1998). The children of war: Street gangs in el salvador. NACLA Report, 21–30.

103 Dell, M. (2015). Trafficking networks and the mexican drug war. The American Economic Journal 105 (6), 1738–1779.

Demombynes, G. (2011). Drug trafficking and violence in central america and beyond. World Development Report 2011 .

DGCP (2018). Estadisticas penitenciarias.

Diario1 (2016). En estos departamentos fueron asesinados ms policas durante 2016. Di- ario1.com.

Does, A. (2013). The Construction of the Maras: Between Politicization and Securitization. Graduate Institute Publications.

Dudley, S. (2010). Drug trafficking organizations in central america: Transportistas, mexican cartels and maras. Woodrow Wilson Center for International Scholars, Latin America Program.

Dudley, S. (2011). Central america besieged: Cartels and maras country threat analysis. Small Wars and Insurgencies, 890–913.

Dudley, S. (2014). Mass graves burying the truth about el salvador gang truce. Insight Crime.

Dudley, S. (2015). Barrio 18 in el salvador: A view from the inside. InSight Crime.

Dudley, S. and E. Pachico (2013). El salvadors gang truce: Positives and negatives. InSight Crime.

Dudley, S. and H. Silva Avalos (2018). Ms13 in the americas: How the worlds most notorious gang defies logic, resists destructio. InSight Crime.

EFE (2018). Centroamrica registr 14,575 homicidios en el 2017, la mayora en 3 pases. El Nuevo Diario.

Espinal-Enriquez, J. and H. Larralde (2015). Analysis of mexicos narco-war network (20072011). PLoS ONE 10 (5).

Focus-Data (2017). 1998, 1999 y 2015 los aos con ms policas asesinados. El Diario de Hoy.

Freedman, E. (2012). El mapa de los medios de comunicacin salvadoreos. Revista En- vio (358).

Garni, A. and L. F. Weyher (2013). Neoliberal mystification. crime and estrangement in el salvador. Sociological Perspectives 56 (4), 623–645.

Garzn, J. C. (2014). Maras y narcotrfico. El Faro.

104 Grillo, I. (2016). Gangster Warlords: Drug Dollars, Killing Fields, and the New Politics of Latin America. Bloomsbury Press.

Gmez Hecht, J. R. (2017). Las pandillas en el salvador: un nuevo tipo de insurgencia? Small Wars Journal.

Holland, A. C. (2013). Right on crime? conservative party politics and mano dura policies in el salvador. Latin American Research Review 48 (1).

Hope, A. (2014). Not quite at war, not yet at peace.

Huezo, M. (2001). Cultura y violencia en el salvador. Violencia en una sociendad en tran- sicion, 115–137.

Huhn, S., A. Oettler, and P. Peetz (2009). Contemporary discourses on violence in central american newspapers. The International Communication Gazette 71 (4), 243–261.

Hume, M. (2007). Mano dura: El salvador responds to gangs. Development in Practice 17 (6), 739–751.

(ICG), I. C. G. (2017). El salvador’s politics of perpetual violence. Latin America Report N64 .

INCSR, I. N. C. S. R. (2014). Volume i drug and chemical contro. Technical report, United States Department of State: Bureau for International Narcotics and Law Enforcement Affairs.

Ingram, M. C. and K. M. Curtis (2014). Homicide in el salvadors municipalities: Spatial clusters and the causal role of neighborhood effects, population pressures, poverty, and education. Wilson Center Latin American Program.

IUDOP (2014). La situacin de la seguridad y la justicia 2009-2014. Instituto Universitario de Opinin Pblica Universidad Centroamericana Jos Simen Caas.

IUDOP, I. U. d. O. P. (2013). Los salvadoreos y salvadoreas evalan el cuarto ao de gobierno de mauricio funes. IUDOP Boletn 2013 No 1 (1).

Justice, D. (2017). Department of justice fact sheet on ms-13. Technical report, Justice Department.

Justice, D. (2018). Violent gang task forces. Technical report, FBI.

Kalyvas, S. (2006). The Logic of Violence in Civil War. Cambridge, MA: Cambridge Uni- versity Press.

Klein, M. W. and C. L. Maxson (2010). Street Gang Patterns and Policies. Oxford University Press.

105 Krebs, V. E. (2001). Mapping networks of terrorist cells. Connections 24, 43–52.

Lacina, B. (2006). Explaining the severity of civil wars. Journal of Conflict Resolution, 276–289.

Lessing, B. (2014). How to build a criminal empire from behind bars: prison gangs and projection of power. Technical report.

Lohmuller, M. (2015). El salvador usa ley antiterrorista para enfrentar pandillas. InSight Crime.

Londoo, J. L. and R. Guerrero (2000). Violencia en amrica latina. epidemiologa y costos. In Asalto al desarrollo. Violencia en Amrica Latina. edited by J. L. Londoo, A. Gaviria and R. Guerrero. Washington, D.C.: Banco Interamericano de Desarrollo.

Lopez, K. (2016). Informes oficiales revelan que existen ms de 100 mil pandilleros. Diario La Hora.

Marroquin, D. (2017). Estos son los rostros de los policas y militares asesinados durante 2017. El Diario de Hoy.

Martin-Baro, I. (1982/1992). Accion e ideologia. Psicologia social desde Centroamerica. San Salvador: UCA Editores.

Martinez, C. and J. L. Sanz (2012). La lista de peticiones que las maras hicieron al gobierno. El Faro.

Martinez, C., J. L. Sanz, and O. Luna (2013). Evolucin de las pandillas en el salvador desde 1945 hasta 2013. Sala Negra.

Martinez, O. (2017). How not to assemble a country. NACLA Report on the Americas 49 (2), 139–144.

Martinez, O., C. Martinez, S. Arauz, and E. Lemus (2012). Gobierno negoci con pandillas reduccin de homicidios. El Faro.

Martinez Barahona, E. and S. Linares Lejarraga (2011). Democracy and punitive populism: exploring the supreme court’s role in el salvador. Democratization 18 (1), 52–74.

Martinez D’aubuisson, J. J. (2015). Ver, oir y callar. Pepitas de Calabaza ed.

Matrix, P. A. (2015). Chapultepec peace agreement. Technical report, Kroc Institute for International Peace Studies, University of Notre Dame.

McAndrew, D. (1999). The structural analysis of criminal networks. In The Social Psy- chology of Crime: Groups, Teams, and Networks. Canter and L. Alison, Eds. Dartmouth Publishing, Aldershot, UK.

106 Molina Vaquerano, F. (2009). Epidemiologia de los homicidios en el salvador. periodo 2001- 2008. Technical report, Instituto de Medicina Legal “Dr. Roberto Masferrer”.

Montoya, A. (2012). The violence of cold war polarities and the fostering of hope: the 2009 elections in postwar el salvador. In Central America in the New Millennium, pp. 49–63. Berghahn Books.

Moodie, E. (2011). The Ethnography of Political Violence: El Salvador in the Aftermath of Peace: Crime, Uncertainty, and the Transition to Democracy. Philadelphia: University of Pennsylvania Press.

Moser, C. and A. Winton (2002). Violence in the central american region. towards an integrated framework of violence reduction. Working Paper 171 .

Osorio, J. (2013). Hobbes on drugs: Understanding drug violence in mexico.

Osorio, J. and A. Reyes (2016). Supervised event coding from text written in spanish: Introducing eventus id. Social Science Computer Review 35, 406–416.

Papachristos, A. V. (2009). Murder by Structure: Dominance Relations and the Social Structure of Gang Homicide 115 (1), 74–128.

Papachristos, A. V., D. M. Hureay, and A. A. Braga (2013). The corner and the crew: The influence of geography and social networks on gang violence. Am Soc Rev 78 (3), 417–447.

Peetz, P. (2008). Discourses on violence in costa rica, el salvador, and nicaragua: Youth, crime, and the responses of the state. GIGA Working Paper No. 80 .

Peetz, P. (2011). Youth violence in central america discourses and policies. Youth and Society 43 (4), 1459–1498.

Perez, O. (2003). Democratic legitimacy and public insecurity: Crime and democracy in el salvador and guatemala. Political Science Quarterly 118 (4), 627–644.

Pitcher, B. L., R. L. Hamblin, and J. L. Miller (1978). The diffusion of collective violence. Am Soc Rev 43 (1), 23–25.

PNC, P. N. C. (2003). Memoria de labores, junio 2002 - mayo 2003.

Quintanilla Avalos, K. V. (2017). Fabricating perceptions of crime: El salvadors mass media and gang repression. Simon Fraser University.

Ramirez Landaverde, M. (2015). Director de la pnc aconseja a policas disparar sin miedo a delincuentes. In El Diario de Hoy. elsalvador.com.

Reuter, P. (2014). Drug markets and organized crime. In Oxford Handbook of Organized Crime, pp. 359–380. Oxford: Oxford University Press.

107 Ribando, C. (2005). Gangs in central america. Congressional Research Service. Ribando Seelke, C. (2011). Gangs in central america. Richani, N. (2010). State capacity in postconflict settings: Explaining criminal violence in el salvador and guatemala. Civil Wars 12 (4), 431–455. Rubio-Fabian, R. (2013). La tregua entre pandillas. La Prensa Grafica. Santacruz, M. and A. Concha-Eastman (2001). Barrio adentro. La solidaridad violenta de las pandillas. San Salvador: IUDOP-UCA/OPS-OMS. Savenije, W. (2009). Maras y bravas: Pandillas y violencia juvenil en los barrios marginales de Centroamrica. San Salvador: FLACSO. Schrodt, P. A. (2010). Automated production of high-volume, real-time political event data. APSA 2010 Annual Meeting Paper. Smutt, M. and L. Miranda (1998). El fenmeno de las pandillas en el salvador san salvador. Technical report, UNICEF/FLACSO Programa El Salvador. Sparrow, M. K. (1991). The application of network analysis to criminal intelligence: An assessment of the prospects. Social Networks 13 (3), 251–274. Storeygard, A., D. Balk, M. Levy, and G. Deane (2008). The global distribution of infant mortality: A subnational spatial view. Population, Space, and Place 14 (3), 209–229. Thale, G., J. Bateman, and A. Goerdt (2013). One year into the gang truce in el salvador: Can the funes administration turn the fragile truce into sustainable public policy? WOLA Commentary. Tilly, C. (1985). War making and state making as organized crime. In Bringing the State Back, edited by P. Evans, D. Rueschemeyer and T. Skocpol. Cambridge: Cambridge Uni- versity Press. Torre, V. d. and A. Martin (2011). Violencia, estado de derecho y polticas punitivas en amrica central. Perfiles Latinoamericanos, No. 37 , 33–50. TRAC, T. R. A. C. (2006). Aggravated felonies and deportation. Technical report, Syracuse Univerisity. TSE, T. S. E. (2018). Resultados de los escrutinios. UNODC (2013). Global study on homicide. UNODC (2014). Homicide statistics 2013. Valencia, R. (2012). El promedio de mungua pays supera en 1.4 homicidios diarios al de melgar. El Faro.

108 Valencia, R. (2014). Las maras no son lo que eran. El Faro.

Van der Borgh, C. (2000). The politics of neoliberalism in postwar el salvador. International Journal of Political Economy 30 (1), 36–54.

Vito, G. F., J. A. Maahs, and R. M. Holmes (2007). Criminology. Theory, Research, and Policy. Sudbury, MA: Jones and Bartlett Publishers.

Wards, T. (2012). Gangsters Without Borders: An Ethnography of a Salvadoran Street Gang. Oxford University Press, London.

Watts, J. (2015). One murder every hour: how el salvador became the homicide capital of the world. The Guardian.

White House, T. (2018). What you need to know about the violent animals of ms-13.

Wolf, S. (2011). Street gangs of el salvador. In T. Bruneau, L. Dammert and E. Skin- ner (Eds.), Maras: gang violence and security in Central America, pp. 43–70. Austin: University of Texas Press.

Wolf, S. (2012). Mano Dura: The Politics of Gang Control in El Salvador. (Austin, TX: University of Texas Press.

Wolf, S. (2017). Mano Dura. The politics of gang control in El Salvador. University of Texas Press. Austin.

Wood, E. J. (2003). Insurgent collective action and civil war in El Salvador. Cambridge University Press.

Xu, J. and H. Chen (2005). Crimenet explorer: a framework for criminal network knowledge discovery. ACM Transactions on Information Systems (TOIS 23 (2), 201–226.

Xu, J., B. Marshall, S. Kazza, and H. Chen (2004). Analyzing and visualizing criminal network dynamics: A case study. Intelligence and Security Informatics, 359–377.

Zeoli, A. M., J. M. Pizarro, S. C. Grady, and C. Melde (2014). Homicide as infectious disease: Using public health methods to investigate the diffusion of homicide. Justice Quarterly 31 (3), 609–632.

ZIlberg, E. (2004). Fools banshed from the kingdom: Remapping geographies of gang vi- olence between the americas (los angeles and san salvador). American Quaterly 3 56, 759–779.

109 BIOGRAPHICAL SKETCH

Viveca Pav´onTercero was born in Tegucigalpa, Honduras. After graduating high school she earned a full athletic scholarship to Meridian Community College in Mississippi. She transfered to Texas A&M University-Commerce to focus on her education. She received a Bachelor of Science in Economics in May 2007 followed by a master’s in Business Admin- istration in 2009. After teaching at various community colleges in the DFW area she was determined to continue to pursue higher education. Viveca was admitted into the Public Policy and Political Economy program at The University of Texas at Dallas in August of 2012.

110 CURRICULUM VITAE

Viveca Pavón

Education University of Texas at Dallas Richardson, Texas Ph.D. Public Policy and Political Economy August 2018 Dissertation: The Dynamics of Violence in El Salvador Advisor: Jennifer S. Holmes M.S. Geospatial Information Sciences August 2018

Texas A&M University – Commerce Commerce, Texas MBA – Minor: Economics May 2009 B.S. Economics May 2007

Publications Peer Reviewed Articles "The Nature of the Honduran Debt and the Possible Impact of its Relief, is it a Vicious Cycle?" E-journal of Business and Economics. of the College of Business at Southern University in LA. Web-publication: http://www.business.subr.edu/index_files/Page6408.htm

Book Chapters 2018. Holmes, J., Pavon, V., and Gutierrez de Piñeres, S. “Economic Development Policies in Colombia (1960s-1990s) and the Turn to Coca in the Andes Amazon.” In The Origins of Cocaine: Colonization and Failed Developments in the Amazon Andies, Gootenberg, P. and Dávalos, L.M. eds. Routledge. Forthcoming.

Manuscripts Under Review “Environmental Provisions in the US-Peru Free Trade Agreement.” (Contributing author with Alisha Kim and Clint Peinhardt) Presented at the Conference on Trade & Environment in Bonn, Germany, in September 2017

“Translating CAMEO: Documenting the semi-automated translation into Spanish.” (Second author with Javier Osorio and Jennifer Holmes)

Works in Progress “The Impact of Gang Networks on Homicides in El Salvador” (Solo author) Article manuscript in progress.

“Modernizing Gold Standard Records (GSR) for event coding of protest data in English and Spanish.” Article manuscript in progress. (First author with Javier Osorio and Jenifer Holmes)

“Violent Lobbying - how powerful are the gangs?” (Solo author) Article manuscript in progress.

“Iron Fist for Votes: the exchange of punitive policies for political gain” (Solo author) Article manuscript in progress. Conference Presentations and Invited Talks 2018. “Efficiency Gains in Focused Geolocation of Spanish Language Political Event Text Analysis” XXXVI International Congress of the Latin American Studies Association (LASA), 2018 Annual Meeting, Barcelona, Spain, May 23-26, 2018

2018. “Efficiency Gains in Focused Geolocation of Spanish Language Political Event Text Analysis” Midwest Political Association (MPSA), 2018 Annual Meeting, Chicago, IL, April 5-8, 2018

2018. “Lessons Learned from the Peace Accords in Colombia and El Salvador” Midwest Political Association (MPSA), 2018 Annual Meeting, Chicago, IL, April 5-8, 2018

2017 “Modernizing Gold Standard Records (GSR) for event coding of protest data in English and Spanish.” Association for Public Policy Analysis and Management (APPAM), November 2-4, 2017.

2017 “Modernizing Gold Standard Records (GSR) for event coding of protest data in English and Spanish.” American Political Science Association (APSA), August 31 – September 3, 2017.

2017 “The Impact of Gang Networks on Homicides in El Salvador” Midwest Political Association (MPSA), 2017 Annual Meeting, Chicago, IL, April 8-9, 2017.

2017 “Translating CAMEO: Documenting the semi-automated translation into Spanish.” Midwest Political Association (MPSA), 2017 Annual Meeting, Chicago, IL, April 8-9, 2017.

2017 Discussant, Movie Screening Roundtable "Finding Oscar" Center for U.S. – Latin America Initiatives, University of Texas at Dallas, April 20, 2017.

2017 “The Impact of Gang Networks on Homicides in El Salvador” Texas Triangle International Relations Conference, Dallas, TX January 28-29, 2017.

2015 “Does Geography Affect Drug Trafficking Crime in Northern Triangle?” (Poster) UTD Annual GIS Day Conference, Richardson, TX, November 18, 2015. Third Place Winner.

2015 Discussant, U.S.-Latin America Special Roundtable "Mexico Today: A Regional Interpretation" Center for U.S. – Latin America Initiatives, University of Texas at Dallas, February 26, 2015.

2008 "The Nature of the Honduran Debt and the Possible Impact of its Relief, is it a Vicious Cycle?" Knowledge Globalization Conference, Boston, MA, April 2008.

Research Experience University of Texas at Dallas Research Assistant for Jennifer S. Holmes (January 2016 – Present) National Science Foundation RIDIR Grant No. SBE-SMA-1539302: Modernizing Political Event Data for Big Data Social Science Research

Teaching Experience USAID Instructor of Record Introduction to GIS (Summer 2017) University of Texas at Dallas Instructor of Record Methods of Quantitative Analysis in the Social and Political Sciences (Fall 2015)

Honors and Awards 2018 Society for Political Methodology annual meeting travel grant 2018 Harold Clarke Dissertation Excellence Award 2017 Center for U.S. Latin America Initiatives (CUSLAI) research grant 2016 Center for U.S. Latin America Initiatives (CUSLAI) research grant 2015 Cyrus Cantrell III UTD Phi Kappa Phi Travel Grant 2015 Phi Kappa Phi Honor Society

Service 2018 Discussant, Midwest Political Science Association, 2018 Annual Meeting 2017 Undergraduate Poster Discussant, Midwest Political Science Association, 2017 Annual Meeting 2017 EPPS Dean’s Student Advisory Council, University of Texas at Dallas 2016 EPPS Dean’s Student Advisory Council, University of Texas at Dallas 2016 Student Evaluator for Incheon National University Exchange Student Presentations 2016 Member, UTD Center for U.S. Latin America Initiatives (CUSLAI) 2015 Member, UTD GIS Student Organization

Technical Skills Statistical Software: R, Stata, BUGS, JAGS Programming and Scripting Languages: Python, SQL Other: ArcGIS, GeoDa, QGis, LaTeX

Languages Spanish – Fluent (Native) English – Fluent