Gang Violence

Concepts, benchmarks, and coding rules

Published on: 10 February 2023 | Last updated: 2 June 2025

Introduction

The Armed Conflict Location & Event Data Project (ACLED) collects data on political violence using a bounded definition of political violence that does not include ordinary crime. The operations of large-scale organized gangs like Mara Salvatrucha (MS-13) in El Salvador, drug cartels in Mexico, and anti-drug operations in the Philippines pose conceptual problems by occupying “a gray zone between ordinary crime and political violence” (Kalyvas, 2015). Political violence can be understood as “the use of force by a group with a political purpose or motivation, or with distinct political effects” according to the ACLED Codebook. Such violence generally manifests as attempts to take over control of a state, make secession demands, increase one’s political influence or position, and/or express demands or views regarding specific political issues. Gang violence in some Latin American countries (e.g. El Salvador, Mexico, and Brazil) and other countries in other regions (e.g. Kenya, Nigeria), however, is not perpetrated with any of these intentions in mind. This seems to render such gang violence ‘non-political’ and sets it apart from more clear-cut examples of political violence seen in these regions.

Yet, large-scale organized crime is not so easily classified as simply ‘criminal.’ Organized crime is not well-defined, but most definitions highlight elements of material or economic benefit and – for larger groups – a notion of monopolizing illegal markets (Reuter, 2015; von Lampe, last accessed on 1 February 2020).¹ While these definitions fit the raison-d’être of groups like MS-13, the Sinaloa Cartel (Mexico), or the Red Command (Brazil), it does not fully capture their range of activities. These groups directly involve themselves in politics through pressure, bribery, directed violence, financing of campaigns, and even grooming cartel members to become politicians. Moreover, some groups have effectively supplanted the state in certain areas by rendering them de facto no-go zones for state forces and instituting state-like structures, such as a judicial system based on their own statutes to hand out penalties to wrongdoers. These indirect yet highly-relevant political acts severely challenge the state’s monopoly of force. The agenda of gangs may not be directly political, but their actions often have a political purpose, motivation, or effect, including the destruction of public order and security. To complicate matters further, these patterns are not always restricted to large and transnational groups; they hold equally true for some national, regional, and sometimes even local gangs.

ACLED’s mission is to understand political violence in its fluid and fast-changing manifestations (OECD, 30 November 2016). Rather than sticking to a rigid definition separating political violence from criminal violence, ACLED aims to capture the disorderly nature of the political manifestations of criminal violence across the globe. This methodology resource intends to tackle the gray area between political and criminal violence by providing conceptual clarifications around when gang violence is political. This document provides specific benchmarks and indicators used to determine 9 of the political, and discusses the complicated issue of attributing violence to unknown actors, who may or may not be gangs. This resource is particularly tailored towards Latin America, but the general approach, benchmarks, and indicators apply equally to other contexts. The term ‘gang violence’ is used to describe violence committed by criminal groups without an overt political agenda. This term is solely meant to communicate that the gangs discussed in this document are organized, have leadership (Spergel, 1995), and that their objective is not political but economic in nature.

When is gang violence political?

ACLED uses a relatively bounded definition of political violence that does not include purely criminal activity. In certain contexts, gang violence may be considered to be more than purely criminal, i.e. when the violence serves a broader political goal or has significant political impacts. Accordingly, ACLED considers gang violence to be political violence and, thereby, includes it in its dataset when the violence is used towards meeting overt political goals. Further, when violent gang activity directly and fundamentally challenges public safety and security, it is also deemed political violence.

Organized crime evolves through gang competition and monopolies over illegal markets. As Klaus von Lampe explains:

“It is true that illegal markets can be linked to high levels of violence. However, not all illegal markets in all countries are equally violent or not all violence that occurs in the context of illegal markets is directed against competitors in an effort to increase market shares and eventually to achieve a monopoly position (Andreas and Wallman, 2009). Violent conflicts can also occur between suppliers and customers, within illegal enterprises, and in the form of predatory attacks on illegal market participants (Amir, 1995; Berg and Loeber, 2015; Wright and Decker, 1997)” (von Lampe, 2016).

Further, ACLED argues that the competition for monopolies and territories involves a perverse incentive to act as a violent governing entity combatting both security services and civilians. It is within this context that ACLED explores the political nature of gang activity.

Many explanations of gang violence point to the political origins of these groups, including political and economic deprivation or failed social policy (Rodgers, 2012; Hagedorn, 2003). While failed social policies and neglect undoubtedly play a role in the emergence of gangs, there are reasons for a more restrictive understanding. The study of political violence is typically concerned with how violence is used as a tool by political actors. ACLED ensures that ‘political intent’ is a key part of defining political violence.

ACLED’s interpretation of gang activity dictates that it qualifies as political violence when it has direct political targets and/or is used for political purposes. For example, when a criminal group or gang targets a (standing or former) mayor, councilor, parliamentarian, minister, or civil servants and their families; the destruction of symbols of public authority (e.g. bombing government property); and assaults on those in the public domain including journalists, human rights activists, and other civil society members. In many cases, this type of violence is employed for a political purpose, and politicians or those with a public interest act as funders, instigators, or benefactors. In the African context, for example, gangs are frequently employed during elections. Across ACLED, this type of gang violence is included for each country and region.

ACLED also includes gang violence when a state’s public security and safety are fundamentally challenged. This happens in places like Haiti and Jamaica, where gangs can act as de facto local governments and public service providers. In the Brazilian favelas, gangs control territory and engage in highly public acts of violence, rendering these places essentially a no-go area for state forces. This violence is neither carried out with an overt political purpose or motivation, nor is it used to that end. However, the decision to include gang violence in these specific circumstances is based on the assumption that the level and nature of gang violence affects and challenges the existing political order. Hence, while the overt purposes and motivations of these groups may not be explicitly political, the consequences for residents and security services are.

When does gang violence fundamentally challenge a state’s public security and safety? Generally, this happens under two conditions. First, when gangs assert some form of de facto control over their territory (e.g. by assuming a role in law and order and/or limiting the state’s entry into the territory, either physically or politically). Second, when gangs challenge the state’s ability to enforce public security and safety by engaging in regular public acts of violence (e.g. battles on the street, public killings, and violence against civilians). These conditions are present in some countries. For example, gang violence in Mexico has rendered state control over some key territories ineffective, and some gangs have taken advantage of crises, like the coronavirus pandemic, to entrench their control in marginalized communities plagued by widespread poverty and lack of access to essential services (for more, see ACLED’s Central America and COVID-19: The Pandemic’s Impact on Gang Violence).

In short, gang violence poses a challenge to the definition of political violence. It is perpetrated by groups that have no overt political motivation. However, some gang violence has clear political connotations and consequences. ACLED considers gang violence as political when it is employed by political actors acting as funders, instigators, or benefactors, and serves identifiable political goals. ACLED also considers gang violence ‘political’ when it fundamentally challenges the public safety, security, and control of the state.

Political gang violence in Latin America

Gang violence that is directly political — e.g. with public figures as targets, investors, instigators, or beneficiaries — is included in every country and region. In addition, as laid out in the preceding sections, ACLED also includes gang violence without such overt political connotations in contexts where it fundamentally challenges public safety, security, and control of the state. This section lays down the benchmarks and indicators used to determine when and where such contexts exist.

Based on key elements identified in the previous section, the determination is made using: a set of general measures about public safety and security; more specific indicators on the extent of de facto control; as well as indicators on the nature of public acts of violence (see Table 1 below).

ACLED’s conclusions about which countries should have their gang violence considered as political were ultimately reviewed by external country experts before being applied systematically in the coding methodology. Similar reviews of countries continue to be done regularly. ACLED researchers collect information weekly in line with real-time coding procedures. Researchers flag cases in which they deem, qualitatively, that a new assessment is warranted for a country (i.e. if its ‘status’ should be changed from a gang exclusion country to inclusion, or vice-versa). In cases where a country previously did not meet inclusion criteria but later does, gang violence in that country is systematically back-coded for the appropriate period before the new gang violence data are published.

Table 1: Indicators to Assess Political Gang Violence

Benchmark	Indicator	Source
Public safety and security at stake	Share of overall violence that is gang violence	ACLED pilot + subsequently qualitative
	Homicide rates	United Nations Office on Drugs and Crime (UNODC)
	Links between politicians and gangs	Qualitative
De facto control	Spatial dispersion of violence	ACLED
	Specialized anti-gang police/military units	Qualitative
	Control of large cities	Qualitative
Public acts of violence	Share of gang violence that targets civilians	ACLED
Public acts of violence	Type and style of public violence	Qualitative

Benchmark 1: Public safety and security

Three indicators are used to determine the general challenge gangs pose to public safety and security directly.

The first indicator is the share of overall violence that is gang violence, based on event data collected by ACLED. To carry out this analysis, ACLED required consistent data for all countries on both political violence as well as gang violence. As existing data could not be found on Latin American gang or political violence in a comparable format, in the lead up to ACLED’s initial expansion to Latin America, ACLED constructed a pilot database containing all reported violence in each country during 2018 and the first half of 2019 (with the exclusion of petty and domestic crime). These event data allowed ACLED to systematically assess the prevalence of gang violence in each country.²

Based on ACLED’s pilot study, the cut-off point for this indicator was set at 10% of overall violence that is gang violence, representing countries with high rates of gang violence. Since this initial assessment, ACLED Researchers flag cases they deem merit a new assessment and this indicator for such cases is reviewed.³⁴The identification of gang violence should be understood as an approximation rather than a definitive statement on gang violence in the country.

The second indicator, homicide rates, uses the average homicide rate for the last five years of available UNODC data.⁵ The indicator shows the number of homicides per 100,000 inhabitants. Based on ACLED’s pilot study, a conservative threshold of 20 homicides per 100,000 inhabitants was set as the cut-off point to determine countries with a high homicide rate.

The third indicator assesses whether there are known links between politicians and gangs (e.g. gang members/leaders holding political positions, financing or provision of services by gangs to politicians and vice versa) and is a qualitative indicator. This was applied to countries with at least one reported gang-related incident. This information is based on publicly-available information collected by ACLED researchers with specialized knowledge on those countries.

Benchmark 2: De facto control

Three indicators are used to assess the extent gangs de facto control territory and challenge state control.

The first indicator assesses the spatial dispersion of gang violence and takes a high level of gang activity that is geographically clustered as an indication that gangs exercise a form of de facto control over territory. This is a quantitative indicator based on ACLED data.

The second indicator assesses the extent to which gangs control large cities, representing centers of power and often includes the capital. This is a qualitative indicator based on ACLED researchers’ knowledge, drawing on information from reports on the extent of geographic control in these urban areas.

The third indicator assesses whether the police and/or military have specialized units for combating gangs. This is meant to capture whether the state itself recognizes gang activity as a particular problem. This is a qualitative indicator for which information comes primarily from official police and military websites and press releases.

Benchmark 3: Public acts of violence

Two indicators are used to assess the extent to which gangs engage in public acts of violence.

The first indicator is the share of gang violence that targets civilians,⁶ based on event data collected by ACLED. This indicator is based on the assumption that a higher share of civilian targeting suggests a purposeful focus on challenging public safety and security through instilling fear in the local population, rather than a focus on engagement with state forces or other armed agents.

The second indicator assesses the type and style of gang violence within a specific country (i.e. the nature and methods of this violence). This indicator is based on a qualitative reading of reports of gang violence in the country, with an aim to capture very ‘public’ violence, such as: daylight attacks; attacks on taxi and bus drivers, government officials, as well as seemingly innocent bystanders; the mutilation of bodies along with their public display; and drive-by shootings. Reports reviewed include general reports on the country and its violence by non-governmental organizations, and both local and international think tanks, as well as violent incident reports used by ACLED researchers for coding purposes. The latter includes over 180 sources in Spanish, Portuguese, French, and English, spanning from local media to international news.

Conclusions

A total of 51 countries and territories⁷ were assessed based on these eight indicators. A country’s gang violence is determined to have a political connotation and, therefore, accepted for inclusion if at least five or more of the eight indicators were satisfied. The results accepted the inclusion of gang violence as political for 11 countries in Latin America of the 51 that were assessed. These countries include Belize, Brazil, El Salvador, Guatemala, Haiti, Honduras, Jamaica, Mexico, Puerto Rico, Trinidad and Tobago, and Venezuela.⁸

Table 2: Summary of gang violence inclusion as political violence

Country	Conclusion	Inclusion
Anguilla	No gang violence reported.	NO
Antigua and Barbuda	No gang violence reported.	NO
Argentina	No inclusion. While public acts of violence exist and civilian targeting is high, all other indicators suggest exclusion.	NO
Aruba	No gang violence reported.	NO
Bahamas	No inclusion. Most indicators either do not have sufficient data to draw robust conclusions, or suggest exclusion.	NO
Barbados	No inclusion. While gang violence is high, specialized police units exist, and gang murders are common, all other indicators suggest exclusion.	NO
Belize	Inclusion. Gang violence is common, clustered, frequently targets civilians and public, and homicide rates are high.	YES
Bolivia	No inclusion. While links between gangs and politicians exist, gang violence mostly targets the government and other indicators suggest exclusion.	NO
Brazil	Inclusion. Nearly all indicators suggest inclusion.	YES
British Virgin Islands	No inclusion. Nearly all indicators suggest exclusion.	NO
Caribbean Netherlands (Bonaire, Sint Eustatius, and Saba)	No gang violence reported.	NO
Cayman Islands	No gang violence reported.	NO
Chile	No gang violence reported.	NO
Colombia	Most indicators are satisfied: homicide rate is high, specialized anti-gang units exist, and gangs have territorial control.	TBD*
Costa Rica	No inclusion. Gang violence involves public killings, but all other indicators suggest exclusion.	NO
Cuba	No gang violence reported.	NO
Curacao	No inclusion. Most data are insufficient to draw robust conclusions.	NO
Dominica	No inclusion. Homicide rate is high, but all other indicators suggest exclusion.	NO
Dominican Republic	Most indicators are satisfied: high share of gang violence, close links, specialized units, territorial control, and lethal violence.	TBD*
Ecuador	Gang violence is common, frequently targets civilians and the public, and homicide rates are high.	YES
El Salvador	Inclusion. Homicide rates are highest in Latin America, gang violence is common and very public, gangs hold control of areas, deep links between gangs and police, and specialized units exist.	YES
Falkland Islands	No gang violence reported.	NO
French Guiana	No inclusion. Most data are insufficient to draw robust conclusions.	NO
Grenada	No inclusion. While the share of gang violence is high, all other indicators suggest exclusion. Little data available in general.	NO
Guadeloupe	No inclusion. Small share of gang violence, low homicide rates and generally little information.	NO
Guatemala	Inclusion. Most indicators suggest inclusions, such as very high homicides rates.	YES
Guyana	No inclusion. Although sufficient data is generally lacking.	NO
Haiti	Inclusion. Most indicators suggest inclusion. Share of total violence is high, many groups control territory and violence is relatively endemic and public. Moreover, there is clear political usage of these groups.	YES
Honduras	Inclusion. All indicators suggest inclusion. Homicides rates are high, there is gang control over territory, and there is highly public violence.	YES
Jamaica	Inclusion. Nearly all indicators are satisfied. Jamaica has one the highest homicide rates in the region. There are clear political links and many small groups engage in common, lethal, and public acts of violence.	YES
Martinique	No inclusion. All indicators suggest exclusion.	NO
Mexico	Inclusion. All indicators suggest inclusion.	YES
Montserrat	No gang violence reported.	NO
Nicaragua	No inclusion. None of the indicators suggests inclusion. Gangs are nearly absent and levels of violence are low.	NO
Panama	No inclusion. Few indicators suggest inclusion. Instances of violence and homicide rates are relatively low. Gang violence occasionally high, but contained within remits of gang feuds.	NO
Paraguay	No inclusion. Most indicators suggest exclusion.	NO
Peru	No inclusion. Almost all indicators indicate exclusion. Share of gang violence is low, low homicide rates, and dispersed violence.	NO
Puerto Rico	Inclusion. Border case as indicators are mixed and sometimes unclear. Homicide rates are high, but bordering exclusion. No clear control of territory. Close links between politicians and gangs, public acts of violence, and high share of gang violence tips balance in favor of inclusion.	YES
Saint Kitts and Nevis	No inclusion. While homicide rates are among the highest in the region, other indicators suggest exclusion. Generally, little data available and status unclear.	NO
Saint Lucia	No inclusion. Little data and status unclear.	NO
Saint-Barthelemy	No gang violence reported.	NO
Saint-Martin	No gang violence reported.	NO
Sint Maarten	No gang violence reported.	NO
South Georgia and the South Sandwich Islands	No gang violence reported.	NO
St. Vincent and Grenadines	No gang violence reported.	NO
Suriname	No inclusion. No indicators supporting inclusion. Generally, little data available and status unclear.	NO
Trinidad and Tobago	Inclusion. Majority of indicators are satisfied: high share of gang violence, high homicide rates, specialized units, links, and type of violence.	YES
Turks and Caicos Islands	No gang violence reported.	NO
Uruguay	No inclusion. All indicators suggest exclusion.	NO
Virgin Islands, U.S.	No inclusion. No indicators supporting inclusion. Generally, little data available and status unclear.	NO
Venezuela	Inclusion. Majority of indicators are satisfied: high share of gang violence, second highest homicide rate in the region, violence is concentrated in large cities, specialized units, links, and type of violence.	YES

* ACLED is currently investigating the inclusion of gang violence as political violence in Colombia and the Dominican Republic.

The attribution of gang violence in Latin America

This final section discusses the problem of when to attribute responsibility for violence by unidentified actors to gangs. ACLED does not use a ‘blanket rule’ by assigning all violence by unidentified actors to gangs in Latin America or in any context. While gangs may be responsible for a significant portion of unattributable violence in countries with a strong gang presence, they do not conduct all of it. This premise has been corroborated through discussions with local partner organizations in the region. For example, a 2016 study on gang violence in Guatemala concluded it could not assign responsibility in 35% of cases where it reviewed homicides in two major hotspots of gang activity, Chiquimula (“trafficking corridor”), and Zona 18 of Guatemala City (“gang area”). In both cases – 28% and 41%, respectively – homicides could be credibly attributed to gangs. The study, therefore, concluded that: “the information from authorities was fragmented, disorganized and sometimes missing altogether” (USAID, 2016). This equally applies to a host of other Latin American, Asian, and African countries, where reporting on violence often does not include the identity of the group involved because it is deliberately concealed by the perpetrator or victim,⁹ self-censorship by the outlet, or simply because no one observed the actual violence, but only its consequences.¹⁰ As such, blanket rules of resorting to gang attribution, even in countries with a strong gang presence, are not appropriate.

Further, there are various incidents for which it is not clear whether the activity is carried out by an individual or an armed group. There are also a large number of other armed groups active in these spaces, often in response to gang violence, who can also be responsible for unidentified violence. And lastly, the demarcation between gangs and other groups is not always clear. These three reasons are briefly discussed below.

Complex actor landscape

A significant problem for the attribution of gang violence is that criminal violence can be perpetrated by an individual rather than an organized armed group (i.e. a gang). For example, an event that reads “On 16 February, a civilian was killed in an armed attack over drugs at Miragoane. Motives are unknown” may have involved an organized group, but may have been perpetrated by an individual. To ensure that only gang violence is included, ACLED only includes events when it is clear that either multiple people were involved in violence (e.g. reports mention “a gang”), or when there is evidence that an individual was acting on the orders of an armed group (e.g. an assassination by a hitman).

Further, there are often various forms of local armed groups present in the same space. For example in Brazil, milícias operate within some of the same neighborhoods as gangs and also contest territory. These milícias were initially formed to ‘protect’ citizens from drug trafficking groups; yet over time, their methods became more coercive as they began charging ‘taxes’ – illegal fees – from residents for ‘protection.’ Typically, these groups fight drug cartels and provide public goods — though they have also become violent and pose a threat to civilians (ALERJ, 2008). Some milícias in Rio state have begun working in drug trafficking themselves, further complicating event attribution (O Globo, 10 October 2019). Moreover, many of these milícias are sponsored by or led by state actors and police, both current and former. In Mexico, communal militias of indigenous groups – so-called Policia Communitaria — were likewise established to protect the community against the activity of gangs. Yet, like in Brazil, police authorities incorporated these self-defense groups into the indigenous justice system, making them state agents in some areas (Oaxaca, San Luis Potosí, and Guerrero). Meanwhile, non-indigenous groups in Mexico have established self-defense forces that have not been integrated into the state. Finally, in countries like Haiti and El Salvador, there are combatants from former paramilitary groups that engage in political as well as criminal violence for various bidders. As a result, the splintered landscape of violence-brokers in countries with active gangs makes it impossible to apply standardized rules across these countries without considering the context.

Finally, even in those contexts in which it is relatively clear that a gang operates, it is not clear to which actor exactly the violence should be attributed. Gangs in Latin America are generally discerned as pandillas and maras. Pandillas are criminal groups organized with a local character (street gangs), they are rooted within a community, and they often provide some level of local public services (Bruneau, 2014). Maras, instead, are nationally or transnationally organized gangs (e.g. MS-13) that often engage in transnational crime and have a different mode of operation. ACLED typically describes armed actors with strong local ties as communal militias – “violent groups that act locally, in the pursuit of local goals, resources, power, and retribution,” according to ACLED’s Definitions of Political Violence, Agents and Event Types. In this way, pandillas are indeed comparable to communal militias in other contexts. While pandillas are still quite common in the urban areas of Haiti or Jamaica, their professionalization in much of Latin America since the 1990s has led to the integration of many pandillas into maras. This factor, in addition to the vagueness of source information, makes distinguishing between pandillas and maras next to impossible, however. As such, gangs in general are treated as akin to political militias given their goals of influencing and impacting governance, security, and policy.¹¹ The difference between groups identifying as pandillas versus maras is quite fluid. A group might identify as one today and the other next month. This makes the distinction around agent categorization (i.e. ‘Interaction’ codes in the ACLED data) less meaningful — and, as such, unlikely to significantly impact analysis.

The various generic actor labels in Latin America

Any actor that is named in the source, including gangs, is coded as such. Hence, an event with notes that read, “On 6 June 2018, the leader of the Compound Gang was killed in a confrontation with the police in Saint Catherine, Jamaica” will be coded as an interaction between the Compound Gang and the state police. However, it is common that violence cannot be attributed to a specific armed group: for example, an event may have notes that read, “On 16 August 2018, an alleged gang member was shot and killed by an unknown armed group In May Pen, Jamaica.” This section defines how ACLED treats these latter cases of various generic actors operating in Latin America, and how they can be discerned from one another and coded.

“Unidentified Gang (Country)”: This label is used for gang violence in countries that qualify for the inclusion of gang violence (outlined in Table 2). This actor is coded when it is clear from reports that a gang is responsible – or it is extremely likely a gang is responsible – for the violence, but the specific gang name (e.g. MS-13) is unknown. Criteria include explicit mention of gang-related violence or some indication of the actor’s criminal nature in the source (e.g. drug trafficking). For some countries, specific modes of violence point to the operation of gangs (e.g. narco message, public hanging, decapitation, drive-by shootings, etc.). As mentioned previously, given that most gangs in Latin America are no longer street gangs but have been integrated into larger gangs, ACLED uses an ‘Interaction’ code of ‘Political Militias’ (3) for all gangs. Gang members are considered to be generally armed on every occasion and, therefore, always coded as “Unidentified Gang (Country),” even if reported as ‘off-duty.’ Hence, the 16 August example above would be coded as an interaction between an “Unidentified Gang (Jamaica)” and an “Unidentified Armed Group (Jamaica).”

“Communal Militia (Country)”: This actor is used when the source indicates that a group is a self-defense group from a particular area, or when this point is clearly inferred from circumstantial evidence. The main differences between communal militias and gangs are apparent in the scope of their membership, actions, and goals. Communal militias only act locally and pursue local goals, resources, power, and retribution. These groups identify as part of a neighborhood, village, or other community. They typically manifest as members of a communal self-defense group against gangs or other armed groups, despite their ability to engage in attacks themselves. Communal militias are coded with an ‘Interaction’ code of ‘Identity Militias’ (4) per ACLED coding methodology. Consider the following example: “On 23 April, a body was found by the police forces in Ciudad Cuauhtemoc, Ciudad de Mexico. A message was left on the body of the victim warning gangs operating in the area and signed by a vigilante group.” This event is coded as a “Communal Militia (Mexico)” that has attacked an “Unidentified Gang (Mexico).”

In Mexico and Brazil, some communal militias have taken on different forms and have started to cooperate with or incorporate into state actors. In these cases, they may be classified either as political militias (coded by ACLED with an ‘Interaction’ code of ‘Political Militias’ (3)), with their specific group name as the actor handle, or as state actors (coded by ACLED with an ‘Interaction’ code of ‘State Forces’ (1)) if fully integrated into the state structure.

“Police Militia”: This actor is specific to Brazil and refers to the milícias¹²– armed groups that are organized and led primarily by current and former police and other state agents. Because of their leadership, they have access to powerful positions within the state and are often more powerful than drug trafficking groups. They were initially formed as a way to ‘protect’ the citizens from drug trafficking groups; however, their methods have since become coercive and they often extort the community in exchange for their protection. An essential characteristic of these milícia groups has been that they are against drug trafficking and contest gang territorial control. However, reports have surfaced describing new alliances between police militias and drug traffickers in Rio state (UOL, 11 January 2019; Instituto Igarapé, 22 May 2018). Sources will name certain police militias (e.g. Liga da Justiça and Chico Bala); most often, however, sources will refer to these groups simply as milícia or miliciano. Given their existence and organization outside of the state, despite their usage by certain state agents, police militias are coded with an ‘Interaction’ code of ‘Political Militias’ (3).¹³ The following is an example event containing police milícias: “On 1 July 2019, in Rio de Janeiro (Rio de Janeiro state), a shootout took place in Morro do 18 (Agua Santa) between a militia group and drug traffickers from the gang Amigos dos Amigos (ADA). During the shootout, several members of the ADA were injured.”

Coding armed groups in Latin America

Coding rules have been developed as a means to determine when to attribute violence to armed actors based on information provided by sources. In countries for which ACLED considers gang violence to be political violence and countries with active insurgencies (see Table 2), ACLED has opted to keep a relatively large amount of violence unattributed to a specific actor; the identity of actors is often hard to verify and the complex environment with a very large number of potential actors as well as their usage of different types of violence complicates attribution. ACLED uses the “Unidentified Armed Group (Country)” label for all these events where attribution is impossible, but the events meet a minimum threshold for inclusion.

Yet there are cases in which it is highly certain that gangs (or other groups) are responsible based on certain tactics, techniques, or procedures (TTPs) or other indicators such as the identity of the victims. Various coding rules are used across different cases where the identity of actors is unclear (discussed below). Unidentified Gangs are generally coded in ‘Battles’ (‘Armed clashes’), in which the source mentions that groups were engaged in drug dealing/trafficking and/or organized crime. For ‘Violence against civilians’ events, except for specific cases in Mexico and Brazil, it is generally unclear who is responsible for attacks against and kidnappings of civilians, and in most cases no responsibility is assumed; some exceptions may include obvious indicators such as narco messages left at the scene.

Coding armed groups in countries where ACLED has deemed gang violence should be included as political violence

In cases where the group or gang is specifically named, the named actor is coded in the event (e.g. Barrio-18). If the source report calls an unnamed group a ‘narco gang’, ‘narco-criminal group’, or ‘drug gang,’ etc., then the unnamed actor is coded as “Unidentified Gang (Country).”

In cases in which violence is carried out by an unidentified armed group, ACLED researchers rely on answers to certain questions to determine whether or not the unnamed actor should be coded as “Unidentified Gang (Country)” or “Unidentified Armed Group (Country)” — or not at all.

In short, such events are only attributed to Unidentified Gang actors if there is mention of drugs (e.g. “two people were shot dead by armed actors; authorities believe the killing is linked to drug-related feuds”) or a narco message.

Otherwise, the event is coded and attributed to “Unidentified Armed Group (Country)” if:

There is evidence that the perpetrator is an armed group (i.e. it is a group of two or more using sophisticated weapons).
If the victim is political. These victims include politicians, government employees, former political actors, public prosecutors, lawyers, judges, other overtly political groups (e.g. human rights defenders), journalists, etc.¹⁴
If there is evidence of the use of TTPs in non-gang-inclusion countries that would be attributed to gangs elsewhere in the region. These TTPs include drive-by shootings, the mutilation of bodies, public hanging of corpses, mass shootings, etc.

In all other cases, the event would not meet the threshold for inclusion at all in the ACLED dataset as political violence, or specifically, gang violence. Examples of such events could be “dead woman found,” “man was killed by an unknown person,” “bones of a man were found,” etc.

Coding of such violence in Mexico, Brazil, and Venezuela,(countries for which ACLED considers gang violence to be political violence), as well as Colombia (a country with an active insurgency), follow additional parameters which are outlined in the respective methodology documents.

Coding armed groups in countries where ACLED has deemed gang violence should not be included as political violence

If the source report calls an unnamed group a ‘narco gang’, ‘narco-criminal group’, or ‘drug gang’, etc., or cites a specific gang by name, then the event is not coded as the violence is deemed to be specifically criminal in countries for which ACLED considers gang violence not to be political violence. This includes cases in which there is mention of drugs or a narco message. However, in cases where the group is specifically named as a transnational cartel (e.g. MS-13, Sinaloa Cartel), the event is coded and is attributed to the group. Such groups are coded when active in their ‘home countries’ – provided the countries are included in the list of gang-inclusion countries – where their activity is considered to meet the political violence threshold, and so continued coverage of such actors is deemed appropriate.

In cases in which violence is carried out by an unnamed armed group, the event is only coded and attributed to “Unidentified Armed Group (Country)” if the victim is political. These victims include those such as politicians, government employees, former political actors, public prosecutors, lawyers, judges, other overtly political groups (e.g. human rights defenders), journalists, etc.¹⁵ In all other cases, the event would not meet the threshold for inclusion in the ACLED dataset.