United States: Coding decisions on actors and unique events

Published on: 28 June 2023 | Last updated: 4 November 2024

Background

The United States continues to be at a heightened risk of political violence and instability following the contentious 2020 general election. Mass shootings recently hit a record high (BBC, 29 December 2019), violent hate crimes are on the rise (Al Jazeera, 13 November 2019), and police killings continue unabated, at 2.5 times the rate for Black men as for White men (FiveThirtyEight, 1 June 2020; Nature, 19 June 2020). The COVID-19 pandemic has killed thousands (New York Times, 3 September 2020) and disrupted the economy, while George Floyd’s death in police custody in late May 2020 sparked a massive wave of protests across the country. The country continues to be polarized, with significant right-wing mobilization. 

The primary tensions in the US center largely around demonstrations and political polarization. 

The US is home to a vibrant demonstration environment characterized by a large proportion of non-violent protests; demonstrations make up the vast majority of events in the US context.  During the summer of 2020, demonstrations associated with the Black Lives Matter (BLM) movement spiked in the aftermath of the killing of Geogre Floyd on 25 May 2020, a result of excessive force in his detainment by Minneapolis police. While overwhelmingly peaceful (for more, see this ACLED report), demonstrations associated with the BLM movement have been met with increased force by authorities, increased engagement by non-state actors, and an increased rate of counter-demonstrators (for more, see this ACLED report). The global coronavirus pandemic has also been a driver of demonstrations. The US has emerged as the new epicenter of the coronavirus outbreak; with only 4% of the world’s population, the US was estimated to have a quarter of confirmed cases by the end of June 2020 (CNN, 30 June 2020). Given not only the economic impact of the pandemic, as well as its politicization, it has fueled a large proportion of demonstrations in the US.

Political violence in the US mostly often involves militias, which have been becoming increasingly proliferate, and which have been engaging in demonstrations at an increasingly high rate; “lone wolves” with political or ideological agendas; and government forces, especially in the form of excessive force used by police, which is often particularly excessive towards select groups (Vox, 14 November 2018). The coding of events in the US has presented a number of methodological challenges for the tracking and recording of political violence and demonstrations. Among these include the coding of demonstrations led by ideological movements rather than by organized groups; an oversaturated media environment; the presence of individual violent actors (a.k.a “lone wolves”) operating independently of an organized group; and the widespread presence of firearms among typically unarmed actors (e.g. Protesters). This document will further expand on ACLED’s decisions concerning these issues and other variables present in the US data.

ACLED’s coverage of the US began with a three-month pilot between July and September 2019. Full data collection commenced in 2020 as part of the US Crisis Monitor, a joint project launched by ACLED and the Bridging Divides Initiative (BDI) at Princeton University to provide the public with real-time data on political violence and demonstrations around the country, establishing an evidence base from which to identify risks, hotspots, and available resources to empower local communities in times of crisis (to learn more about the original 2020 project, see the archived project page. In 2024, ACLED re-launched a new, independent version of the US Crisis Monitor to offer dedicated analysis in the lead-up to the 2024 election.

How does ACLED code certain actors in the United States?

A number of unique actors are present in the US data, among them sub-divisions of security forces, a number of militias, and unique associated actors to civilians and demonstrators.

Security Forces of the United States

ACLED codes specific sub-groups of state forces in cases when it is deemed analytically useful. Normally within ACLED coverage, this occurs in regions where a) the sources consistently differentiate between sub-groups; and b) the sub-groups engage in distinct types of events (e.g. terrorism task forces, gang violence units, military police, etc.). A non-exhaustive list of armed state actors in the US includes:

Actor Name	Details
Military Forces of the United States (2021-)	Used for all actors simply referred to as “military forces”, as well as all other units under the administration of the Department of Defense which have not been determined analytically useful to have as separate actors (see below).
Military Forces of the United States (2021-) National Guard	Includes all state National Guard units.
Police Forces of the United States (2021-)	Used for all municipal and state police units not captured by the more specific police actors below. Also used when the source mentions only a general “police” or “security forces” without any further indicators.
Police Forces of the United States (2021-) Border Patrol	Includes all police units within the Customs and Border Protection agency. This is the default police actor used for events involving migrants or refugees along the US border, unless otherwise reported.
Police Forces of the United States (2021-) Immigration and Customs Enforcement	Used for police units operating under ICE (Immigration and Customs Enforcement), a sister organization to the Border Patrol but operating under the Department of Homeland Security (DHS). This is the default actor for any events involving migrants away from the border, unless otherwise noted by the source.
Police Forces of the United States (2021-) Federal Bureau of Investigation	Used for all FBI police units.
Police Forces of the United States (2021-) Prison Guards	Used to capture “prison guards” and any other units operating within prisons, unless otherwise stated by the source. This is the default police actor for any event occurring within a prison.
Police Forces of the United States (2021-) Secret Service	Used to capture police units assigned to protect high-level US officials such as the President of the United States, visiting foreign heads, etc.
Police Forces of the United States (2021-) Protecting American Communities Task Force	Used to capture agents operating as part of the PACT task force created by the Department of Homeland Security on the Executive Order of the US President on 1 July 2020. Their mandate is to “protect monuments, memorials and statues”, and the order allows for federal officers to be deployed without the permission of individual US states. While ACLED may have individual actors for the units operating under this task force, the vagueness of reports in identifying the specific units involved in events has made the use of a single  “task force” actor a more viable solution. An incomplete list of units which may operate as part of the PACT force include, but are not limited to: the Federal Protective Service, the United States Secret Service, the Immigration and Customs Enforcement, Customs and Border Protection agency, and the Transportation Security Administration.
Police Forces of the United States (2021-) Operation Legend	Similarly to PACT, Operation Legend captures agents operating under the so-named federal task force announced on 8 June 2020 to help combat crime, initially in Kansas City. An non-extensive list of units which may operate under Operation Legend include: the United States Department of Justice; the Federal Bureau of Investigation; the United States Marshals Service; the Department of Homeland Security; the Drug Enforcement Administration; and the Bureau of Alcohol, Tobacco, Firearms and Explosives.

Across all regions of ACLED coverage, ACLED adds ‘regime/administration years’ (in parentheses) to the end of state actor names in order to designate the specific regime/administration under which those forces acted. Regime/administration years featuring a date followed by a dash denote state actors operating under the current regime/administration. Once a new regime/administration comes into power, the dash is followed by the year in which the previous regime/administration ended its tenure; all prior events will then be updated to reflect this change. Denoting as such helps users analyze differences in trends of activity of state forces under different regimes/administrations.

In the US context, state actors include “(2017-2021)” to denote state forces active under the Trump administration. Events featuring state forces from 20 January 2021 onward are considered to be under the Biden administration, and thus their regime years appear as (2021-). If the Biden administration is succeeded by a new administration in 2025, state actors active during the Biden administration will be updated to include “(2021-2025)” following their names while those active under the new administration will include “(2025-)”; this will allow users to track any differences in the trends of activity of state forces across different administrations. This is done across all countries of ACLED coverage.

Non-state armed actors

Non-state actors in the US data include local “militia” groups, members of right-wing or left-wing armed groups, and “lone wolves” engaging in political violence.

Of the non-state groups, local and state militias make up the majority of armed actors appearing in the US data. These are armed groups operating in both urban and rural contexts, many of whom consider themselves to be local security providers. These militias are mostly community based and aim to defend their local community; therefore, such entities are coded as Unidentified Communal Militia (United States) if they are not affiliated with a specific named group. If the group in question is named, the specific name of that group is instead coded (e.g. New Mexico Civil Guard). All communal militias in the primary actor columns can be found by filtering for ‘Identity Militias’ or its corresponding numeric value, 4, in the appropriate Inter column. For more on interaction terms, see the ACLED Codebook.

Similarly, the Unidentified Armed Group (United States) actor is used for reports of unidentified armed persons who are suspected to be members of armed groups with primarily political agendas (e.g. Boogaloo Boys). These groups are often referred to by the sources as being either radical right-wing or left-wing groups. If the group in question is named, the specific name of that group is instead coded (e.g. KKK: Klu Klux Klan). All political militias in primary actor columns can be found using the appropriate Inter column, and filtering on ‘Political Militias’ (3). For more on interaction terms, see the ACLED Codebook.

Lastly, unique to the US context within ACLED coding is activity carried out by a single person, often described as a “lone wolf,” without a clear affiliation to a specific named group, coded as Sole Perpetrator (United States). This actor is introduced to the data in the US context, where the most lethal cases of political violence are mass shootings often carried out by a sole perpetrator. Hence, attributing such shootings to an ‘Unidentified Armed Group’ would not be accurate given the lone perpetrator of such events. Despite being unaffiliated to any group, these sole perpetrators frequently connect to a kind of political subculture or social movement that motivates their actions, making the tracking of their activity important. The actor is used when an individual is not clearly part of a group and acts alone, such as in the case of mass attacks (e.g. mass shootings, “lone wolf” bombings, car ramming attacks on crowds, etc.) or politically-motivated attacks.

Splinter and affiliate militia groups

For larger, national militia groups such as the Three Percenters, which have a number of affiliates and splinter groups, the different groups are coded as separate actors but will maintain a connecting word so that they can be filtered together. In the case of the Three Percenters, all actors affiliated with, or splintered from (while still maintaining the III% identity), will have the III% term in the actor name. For example, the following is a list of some Three Percenter affiliates and splinters appearing as actors in the data:

• III% Originals
• III% American Patriots
• III% Security Force
• III% We the People

In many cases, the specific affiliate or splinter group of the Three Percenters  will not be mentioned by the source, in which case the actor Three Percenters (III%) will be used.

Subnational chapters of the above groups are not coded as separate actors, but rather will be mentioned in the Notes column (e.g. Georgia III%). Some exemptions to this rule may be made in cases where the subnational chapter operates very independently from the larger movement, such as the III% Washington State.

Unique associated actors

There are cases where the actors in the ‘associated actors’ category require further explanation. 

Movement actors:

The BLM: Black Lives Matter is an associated actor used to define demonstrators associated with the broad social movement; it is coded as an associated actor in the data (most often to Protesters or Rioters) across all regions around the world under the following instances:

• when the demonstration has a local or national BLM group involved; 
• when the main issue of the demonstration concerns the police killing of a specific Black person (e.g. George Floyd, Breonna Taylor); 
• when the main issue of the demonstration concerns police brutality against Black people in general; or 
• when the demonstration is in solidarity with the movement in the US against police brutality against Black people. 

Given the above, it is important to note that the coding of “BLM: Black Lives Matter” as an associated actor in the data is not meant to suggest that all associated events are directly affiliated with the national BLM organization; rather, it notates an association specifically, along the above-outlined parameters.

ACLED does not code BLM as an associated actor when: 

• the demonstration concerns police brutality in general (without a racial element); 
• when the demonstration concerns police brutality against another identity group (e.g. Latin Americans); 
• when the demonstration concerns the police killing of a specific person who is not Black; or 
• when the demonstration concerns other issues related to racial tensions or racial inequality. 

While the above demonstrations are indeed coded, especially as they relate to social justice, they are not specifically coded as associated to the BLM movement unless they meet the initial criteria presented above.

Similarly, a number of other associated actors are used in this way, capturing both the presence of members of a specific activist group, in addition to those demanding something as part of a broader movement by the same name. These include, but are not limited to:

• Cancel the Rents: used to capture the broader movement of demonstrations demanding a cancellation or temporary suspension of rent during the coronavirus pandemic, in addition to those events specifically organized by the activist organization of the same name.
• Abolish ICE: used to capture the broader movement of demonstrations demanding the abolition of the US Immigration and Customs Enforcement (ICE) federal law enforcement agency, in addition to those events specifically organized by the activist organization of the same name.
• Save our Children: used to capture the broader movement of demonstrations associated with the online hashtag and activist movement Save our Children, protesting child trafficking in general, and in some cases related to the QAnon conspiracy theory. For demonstrations associated with Save our Children in which there is specifically some evidence of support for QAnon, the QAnon associated actor will also be added to the event.
• Stop Asian Hate: used to capture the broader movement of demonstrations demanding an end to prejudice and racist attacks against members of the Asian and/or Pacific Islander community. Events which are known to be led by, or featuring an organized group of, members of the Asian/Pacific Islander community will also feature the Asian American Group (United States) associated actor.

Racial and ethnic identity groups:

The African American Group (United States) associated actor is used for groups or individuals whose salient identity is African American. It is not used in association with the BLM movement by default as BLM is a movement actor and many demonstrations are comprised of numerous other races and ethnicities.

Similarly, the associated actors Latin American Group (United States) and Asian American Group (United States) are coded for any group or individual whose salient identity is Latin American or Asian American. 

For associated actors based on race or ethnicity, such as African Americans or Latin Americans, it is important to note that the inclusion of the associated actor is not an attempt to racially profile the attendees of a demonstration. Rather, their inclusion is to note the importance of that racial/ethnic identity as a fundamental component to the organization of a specific demonstration. These associated actors will most often (but not exclusively) be used under two circumstances: 

1. The demonstration is specifically organized by members of a racial or ethnic community in order to address issues specific to that community (e.g. “members of the Black community demonstrated in Minneapolis (Minnesota) against removing the city’s first Black police chief.”)
2. The demonstration is organized by, or features, a specific group or organization in which race or ethnicity is an integral component (e.g. events featuring members of the NAACP: National Association for the Advancement of Colored People will have both NAACP and African American Group as associated actors).

Other unique associated actors:

The Pro-Police Group (United States) is a general associated actor used to capture demonstrations which are held in support of police. The specific movement groups Blue Lives Matter or Back the Blue are added in addition to this actor if mentioned explicitly by the source. For example, if a source says “people from the group Blue Lives Matter gathered to demonstrate in support of the police” the event would have both Blue Lives Matter and “Pro-Police Group (United States)” added as associated actors.

White Nationalists (United States) is coded as an associated actor to actors – including Protesters, Rioters, Unidentified Communal Militias, or Sole Perpetrators – if they are specifically referred to as white nationalists, or if the event context makes this association clear (e.g. the event is a lynching or attempted lynching of a person of color, white nationalist writing or symbols are present, or the perpetrators refer to themselves as Nazis, etc.). In instances where a group might have ties to white nationalism, then the actor will appear as that specific group name (e.g. NSM: National Socialist Movement). The White Nationalists associated actor is only used in cases when no named group is referenced in reporting. Whether or not named groups identify as ‘white nationalist’ or others view them as such is more subjective, and is hence not coded in the data. For those wishing to explore trends around white nationalism, they should review events with the White Nationalists associated actor as well as events involving all named groups that the user believes to be white nationalist; the latter is an analytical decision that the user must make. 

How does ACLED code certain unique events in the United States?

ACLED codes a number of tags for event trends that are unique to the United States. This section discusses the tags that are coded exclusively for events in the United States. For the full list of tags coded in the ACLED dataset, including tags that are coded for events in all countries, see Tags in ACLED data.

Presence of firearms at demonstrations

Demonstrators are, per ACLED definition, either protesters who do not engage in violence or they are spontaneously organized rioters who engage in violence or destructive behavior in the context of a demonstration, typically while unarmed or crudely armed. However, on occasion, demonstrators can be armed with more sophisticated weapons in certain contexts. In the US in particular, demonstrators at times appear with firearms, especially given the prevalence of arms in the country; under these circumstances, their actions dictate how they are coded – whether they engage in violence by making use of their weapons during the demonstration, or not. 

If “armed protesters” engage in a demonstration without physical violence, and do not appear to be part of any organized group, they are coded as Protesters (with an Inter code of ‘Protesters’ [6]). For example: “roughly 1000 people staged a Freedom rally in Harrisburg (Pennsylvania) against the decision by the Governor to close bars and restaurants amid the coronavirus pandemic. Some people were armed at the rally and refused to wear masks while attending”.

In other cases, armed individuals may be present at a demonstration, yet are not themselves involved in the demonstration as a demonstrator, and do not engage physically with demonstrators. Often these reports cite the armed presence as locals present to “defend their community” from rioting, or to intimidate demonstrators. Such demonstrations will include a tag “armed presence” in the Tags column of the event to denote the presence of such armed individuals, as these armed actors are not otherwise coded as actors in the demonstration event, since they did not directly engage in the event. Additionally, the armed incident — separate from the demonstration in which the armed agents did not engage — is coded separately with event type “Strategic developments”, sub-event type “Other”. The Notes for these events will start with “Non-violent activity:”. For more on “Strategic developments”, see below. For example: “more than 1,000 people staged a peace vigil and demonstration in Reno (Nevada) in support of the Black Lives Matter movement. An unreported number of armed people belonging to the Light Foot Militia, the Boogaloo Movement, and a motorcycle club Lucky Ones Nevada were present and claimed to be there to ‘keep the peace’ yet did not directly engage with the protesters. .” The Tags column for this event will have the tag “armed presence”.

All demonstration events in which firearms are present, whether those armed are participants in the demonstration (hence coded as an actor in the demonstration event) or are present yet do not engage in the demonstration (hence not coded as an actor in the demonstration event, though noted in the Notes), will include a tag “armed” in the Tags column to denote the presence of firearms at the demonstration. The “tag” will only appear in the Tags column of demonstration events and will not be included in other Event Types. For example: “Armed demonstrators from the Boogaloo-associated Virginia Kekoas and BLM757 gathered to protest in downtown Richmond (Virginia), as they attempted to walk up to the Virginia Capitol with their weapons. They were protesting against new gun laws that restrict firearms from state grounds and were turned away by Capitol police. They relocated a few blocks away and continued their demonstration.” The Tags column for this event will have the tag “armed”.

Presence of militias at demonstrations

If demonstrators are seemingly part of an organized militia group (named or otherwise) yet they do not use physical violence, the associated actor to Protesters is coded as either Unidentified Communal Militia, if the specific group name is not mentioned, or as the specific group name, if known (e.g. Proud Boys). For example: “protesters gathered in Lansing (Michigan) to protest against gun control measures. Among the groups present were the Michigan Home Guard, the Michigan Liberty Militia, and the Proud Boys, all of whom showed armed and wearing plate carriers”.

On rare occasions, demonstrators who are members of a militia may engage in “low-scale” physical violence (e.g. pushing, punching, spitting, etc.) yet still remain “demonstrators” within that context (i.e. the primary actor is coded as Rioters, rather than as a militia, and the militia actor is coded as associated actor to Rioters). For example: “a group of 50 people, among them the No Sleep Till Justice DFW Group, some armed, held a demonstration in Weatherford (Texas) calling for the removal of a Confederate monument located at the county courthouse grounds. The group was met by an estimated 500 defenders of the statue, including armed members of the III% American Patriot, who threw bottles and other objects at them. This led to physical scuffles between both groups… Neither injuries nor arrests were reported”.

If militia members engage violently with demonstrators using their weapons, or in any other large-scale and organized fashion, they are coded as a militia (by name if known, or as an Unidentified Armed Group or Unidentified Communal Militia, depending on their goals, if name is unknown) (i.e. with the Inter code Political Militias (3) or Identity Militias (4), respectively) rather than as “demonstrators”. For example: “people staged a demonstration in Albuquerque (New Mexico) in support of the Black Lives Matter movement. The demonstration turned violent when a group of people attempted to damage La Jornada sculpture, which they wanted to remove. Members of the New Mexico Civil Guard clashed with demonstrators, with one of the latter eventually shooting at an individual”.

Car rammings and other attacks during demonstrations

Car rammings against peaceful protesters are coded as “Excessive force against protesters” given the propensity for such attacks to be lethal. If a single person drives their car into demonstrators, the primary actor is coded as Sole Perpetrator (United States) against Protesters, unless the actor belongs to a specific mentioned group (e.g. the KKK: Klu Klux Klan). Car ramming events include a tag “car ramming” in the Tags column. ACLED only codes events as “Excessive force against protesters” resulting from car ramming when a car was driven into protesters on purpose, or is under investigation due to suspicion of purposeful intent (with the event updated if/when new information comes to light). This means that events in which a car accidentally hits protesters (e.g. a car is surrounded at an intersection, slowly rolling forward) are not coded as “Excessive force against protesters”; because such events are not intentional, the driver is not driving with enough force to cause serious injury, and there are therefore no serious casualties reported (in the event of serious casualties, the event would indeed be coded as “Excessive force against protesters”). If a car is driven into demonstrators who are engaging in violence or destructive activity (e.g. Rioters), the event is coded as “Violent demonstration”, rather than “Excessive force against protesters”, and the same metrics mentioned above (i.e. the driver’s intent and/or serious casualties) are used to determine whether the event should qualify for a second actor (as above, Sole Perpetrator or the specific group) and the “car ramming” tag. 

Similarly, ACLED codes other attacks by outsiders during demonstrations as either  “Excessive force against protesters” or “Violent demonstration”, depending on whether or not the targeted demonstrators are engaging in any sort of destructive or disruptive activity (i.e. whether they are coded as Protesters, or Rioters, respectively). For an attack against peaceful protesters to be deemed “excessive force”, the violence must be either lethal, have the propensity to be lethal, or result in serious injuries/hospitalization.

Suggested ‘infiltrators’

Reports of infiltrations in demonstrations are very difficult to verify, as demonstrators may themselves be violent or destructive for various reasons. ACLED relies on its sources to investigate and report on instances of suspected infiltrations, and only codes such events when the evidence is sufficient. 

• If an infiltration is reported within a demonstration, and the infiltrators are the only individuals engaging in violence, then the event is coded as a separate event (a “Violent demonstration”) from the “Peaceful protest” (or “Protest with intervention”; “Excessive force against protesters”). For example, the following would be coded as two events, one “Peaceful protest”, and one “Violent demonstration”: “a small number of people engaged in rioting in Lancaster (Pennsylvania), separate from a peaceful protest held outside the police station against the killing of Ricardo Munoz. The rioters damaged lights outside the Post Office, tried to set a tree on fire, and threw rocks at the police station and a police vehicle. 2 people were arrested, one of them being from out of town.“
• Often, in cases where infiltrators are reported, they serve as instigators of violence, and an initially peaceful protest turns violent. Such cases will be coded as a single “Violent demonstration” – even if the violence was instigated by an infiltrator, given that (some) demonstrators themselves engaged in violence, vandalism, looting, etc. as well. For example: “demonstrators looted and vandalized businesses in Naperville (Illinois) as part of the Black Lives Matter movement. The police arrested one person who threw fireworks at police and 11 looters were also arrested. The local police chief told reporters some of the looters came from outside Naperville and used bricks, two-by-fours and bottles to shatter store windows”.

Depending on the information provided by the source on the identity of the individual(s), the infiltrating actor is coded as either a Sole Perpetrator (for lone-wolf style attacks); an Unidentified Communal Militia (for reports of local militia groups); an Unidentified Armed Group (for reports of unidentified armed persons, or suspected members of radical right-wing or left-wing groups); as the specific group name, if known (e.g. Boogaloo Bois); or as Rioters, if the perpetrators are unorganized and of unknown affiliation (e.g. reports of looters coming from “out of town”).

Events in which reports suggest such instigators were present are denoted with a tag “suggested agents provocateurs” in the Tags column of the event. The “suggested agents provocateurs” tag is applied to events from 1 January 2020 to 31 December 2020 in the data.

Detentions by state forces in unmarked vehicles

In July 2020 reports surfaced that federal law enforcement officers emerged from unmarked vehicles and detained protesters without a warrant or further explanations. Events in which demonstrators, or other civilians, are taken and detained by state forces in unmarked vehicles outside of demonstrations are coded with event type “Strategic developments”, sub-event type “Arrests” and are denoted with a tag “detentions” in the Tags column of the event. The “detentions” tag is applied to events from 1 January 2020 to 31 December 2020 in the data. If the detention occurs during a demonstration, the tag will be added to the specific demonstration event during which it occurred. These events are not limited to federal forces, but rather any state security force which meets the above parameters.

Attacks on statues and monuments

If political statues (e.g. Confederate leaders, Christopher Columbus, etc.) are targeted in the context of a riot (i.e. by actors not part of an organized armed group), the event is coded as event type “Riots”, sub-event type “Violent demonstration”, with the primary actor coded as Rioters. If a statue is targeted during demonstrations in which rioters also engage with police, then police are coded as the second primary actor. 

For ACLED to code attacks on statues as a “Violent demonstration”, there must be substantial damage to the statue – rather than simply graffiti or other similar acts of minor vandalism – such as the breaking off of pieces or the toppling of the statue. 

Events in which statues are targeted are marked with a tag “statue”in the Tags column. The “statue” tag is applied to events from 1 January 2020 to 31 December 2020 in the data.

How does ACLED code counter-demonstrations and violence against journalists and health workers present at demonstrations?

Counter-demonstrations

If two protests occur at the same time, in the same place, and they are essentially “counter” to one another – for example, one is pro-mask mandate, and the other is anti-mask mandate – the events are coded as a single event instead of as two separate events. Depending on whether each side engages in destructive or disruptive behavior, each side is coded as Rioters or as Protesters. As such, the Interaction code for such events can be Rioters-Rioters (55), Rioters-Protesters (56), or Protesters-Protesters (66). If both sides engage in violence, the event is coded as a “Violent demonstration”; if only one side engages in violence against the other, the event is coded as a “Protest with intervention” or as “Excessive force against protesters”, depending on the degree of violence; and if neither side engages in violence , the event is coded as a “Peaceful protest”. All counter-demonstrations will have a tag “counter-demonstration” included in the Tags column of the event.

In some cases, police may also engage in such a demonstration, engaging with one or both groups of demonstrators. Such events may be coded with an Interaction code of State forces-Rioters (15) or State forces-Protesters (16), if the interaction between demonstrators and police is deemed more significant than the interaction between the demonstrators. In such cases, a tag “counter-demonstration” will be included in the Tags column of the event. 

To identify all counter-demonstrations, users should look for events with “counter-demonstration” in the Tags column.

Violence against journalists and health workers during demonstrations

The targeting of journalists and/or health workers often occurs within the context of demonstrations that they may have been trying to cover/assist in. Thus, these events are most often contained within the demonstration events themselves, as coding them separately (as a “Violence against civilians” event, for example) would lead to double-counting of the same event. As such, this targeting will appear within the context of the demonstration event, and will list both Journalists or Health Workers and Civilians as an associated actor. Events listing Journalists or Health Workers alone as an associated actor to Protesters or Rioters refer to events in which journalists or health workers themselves may be demonstrating (e.g. nurses demonstrating against a decision to lift coronavirus restrictions). For further information on ACLED events involving journalists and media workers, see this primer. Similarly, this primer provides information on events involving health workers.

What types of strategic developments does ACLED capture in the United States?

The ”Strategic development” event type is unique from other event types in the ACLED dataset in that it captures significant “developments”. Because the types of events which may be considered significant varies by context as well as over time, these events are, by definition, not systematically coded. One action may be significant in one place at a specific time yet a similar action in a different context during a different time period might not have the same significance. This means that “Strategic developments” should not be assumed to be cross-context and time comparable like other ACLED event types are. Rather, “Strategic developments” ought to be used as a means to better understand analysis you are conducting as a user. For more on “Strategic developments”, see this methodology primer.

For the US, “Strategic developments” in the data include, but are not limited to:

• The presence of  an armed group or militia at a demonstration, which does not engage physically with any other groups and does not join the demonstration. These occurrences are coded as a separate event from the demonstration event, with event type “Strategic developments”, sub-event type “Other” and the Notes will begin with “Non-violent activity:” to denote as such. For more, see the Presence of Militias at Demonstrations section above.
• The deployment or movement of security forces are coded with event type “Strategic developments”, sub-event type “Change to group/activity”. The Notes will begin with “Movement of forces:” to denote as such. This includes all deployments of federal troops such as the National Guard or members of specific Department of Homeland Security (DHS) task forces (e.g. Protecting American Communities Task Force [PACT]; Operation Legend).
• Security measures enacted by the state are coded with event type “Strategic developments”, sub-event type “Change to group/activity”. The Notes will begin with “Security measures:” to denote as such. This includes, but is not limited to: curfews, the prohibition of gatherings, drone surveillance of demonstrations, and the creation of checkpoints.
• The presence of armed groups operating patrols or gathered as a means of community ‘protection’ in anticipation of riots or demonstrations that do not materialize is coded similarly to how demonstrations featuring an armed presence ‘tag’ are coded, with event type “Strategic developments”, sub-event type “Other”, and “Non-violent activity:” at the beginning of the Notes.
• Targeted attacks on the property of political targets in which significant physical damage is done (e.g. windows/doors/furniture smashed, documents/computers destroyed, etc.) will have the event type coded as “Strategic developments”, sub-event type “Looting/property destruction”. The Notes will begin with “Property destruction:” to denote as such. For example: “Property destruction: On 10 October 2020, a man threw multiple Molotov cocktails at a Planned Parenthood facility in Fort Myers (Florida). The perpetrator was charged with first-degree arson.”

How are locations coded in the United States?

ACLED provides up to three administrative divisions for each of its country datasets. Typically, these levels are based off of official administrative boundaries. In some cases, official boundaries may not exist for higher level divisions. In other cases, ACLED may forgo including higher level divisions if they are deemed to be unuseful, such as in the US, where the Admin 3 divisions are essentially the same as Location names in many cases. Below are the administrative divisions included in the data for the US.

ADMIN 1: State or Washington D.C.

ADMIN 2: County

ADMIN 3: Not Applicable (ACLED does not code ADMIN 3 in the US)

LOCATION: A populated place (city, village, etc.), natural landmark (hill or mountain, bay, etc.), or a distinct location outside the borders of a population center (military bases, rural airports, etc.)

Map 1: First Order Admin Divisions of the United States ¹ Map courtesy of Wikimedia Commons

How does ACLED code locations which straddle multiple counties?

For locations which straddle multiple counties, such as Austin, Texas – split between Travis, Hays and Williamson counties – the county in which the town hall or state capitol is located is chosen as the Admin 2 (i.e. Travis county for Austin, Texas).

Are there locations in the United States coded below the city level?

ACLED’s standard procedure for coding populated places such as cities, towns, or villages is to code a single location with central coordinates for each. This method accounts for the variable levels of specificity across sources, which may or may not provide information on the exact location of an event within a population center (e.g. neighborhoods). As a result, in the vast majority of cases ACLED location data cannot be used to identify the precise location of an event within a populated area, since the granularity of the data does not exceed the city/town/village level (e.g. if an event is reported by one source to have occurred at the intersection of Third and Broad street in Richmond (Virginia), the coordinates for the event would instead be a single set of centrally located coordinates used for all events taking place in Richmond – in this case 37.5388, -77.4336). However, there are rare exceptions to this rule in situations where coding a sub-region of a city is both possible and analytically useful. This exception applies only to a select few large cities (usually capitals), cities in which events like sieges may be better captured using sub-regions (e.g. the recapturing of Mosul in Iraq), or sub-regions within cities which are temporarily autonomous (such as protest zones or occupied neighborhoods). In the United States, there are five cities in which certain sub-regions are captured by unique locations coded below the city-level.

The cities that are coded below the city-level are Chicago, Illinois; Los Angeles, California; New York City, New York; Portland, Oregon; and Washington, DC. In cases in which the source is unspecific about the location of the event within the city, the event is coded at the general city location. In all other cases, where more detailed information is available, the more specific sub-city location is coded. See below for a complete list of the subdivisions for the cities:

Chicago is coded into nine “sides”:

• Chicago – Central
• Chicago – Far North Side
• Chicago – Far Southeast Side
• Chicago – Far Southwest Side
• Chicago – North Side
• Chicago – Northwest Side
• Chicago – South Side
• Chicago – Southwest Side
• Chicago – West Side

Los Angeles is coded into seven regions:

• Los Angeles – Central
• Los Angeles – Eastside
• Los Angeles – Harbor
• Los Angeles – Northeast
• Los Angeles – San Fernando Valley
• Los Angeles – South
• Los Angeles – Westside

New York City is coded into five boroughs:

• New York – Bronx
• New York – Brooklyn
• New York – Manhattan
• New York – Queens
• New York – Staten Island

Portland is coded into seven regions:

• Portland – Downtown
• Portland – North
• Portland – Northeast
• Portland – Northwest
• Portland – South
• Portland – Southeast
• Portland – Southwest

And Washington, DC is coded into four quadrants and the National Mall:

• Washington DC – National Mall
• Washington DC – Northeast
• Washington DC – Northwest
• Washington DC – Southeast
• Washington DC – Southwest

In addition to these five subdivided cities, the temporary Capitol Hill Organized Protest (CHOP) zone appears in the data as a separate location in Seattle between 8 June 2020 (when it was established) and 1 July 2020 (when it was cleared by police). This location’s name is stylized as Seattle – CHOP.

How are the Unincorporated Territories of the United States coded?

The Unincorporated Territories of the United States are coded within their specific geographic regions for the sake of capturing regional context. For this reason, both Puerto Rico and the US Virgin Islands are coded within the Caribbean, rather than North America. Likewise, American Samoa, Guam, and the Northern Mariana Islands are coded within Oceania.. For all of the above, the territory name will occupy the Country column, rather than the Admin1 column for ease of analysis. 

How are events sourced for the United States?

ACLED researchers review over 2,800 sources to code political violence, demonstrations, and strategic developments across the US. As in other regions, the creation of a balanced source list is a primary focus. While there is evidence that press freedoms have declined since 2016 (RSF USA, 2020), media coverage in the US remains varied and widespread. In truth, an oversaturation of media coverage in the US has posed the most significant challenges. National, regional, and international sources often focus heavily on high-profile events and regions, such as the Northeast and California. Smaller-scale and more local events are instead captured by subnational sources, both at the state and county/municipal level. In practice, this means that thousands of sources require review in order to capture the majority of events across the country, especially in more rural areas or lower population states, like in the Midwest. The results of this endeavor to equalize coverage across states is visualized in Figure 1 below, which outlines the percentage of events that reference a certain type of source (using the Source Scale variable) by State.

Moreover, beyond creating a balanced source list in the endeavor for equal geographic coverage, the political spectrum of sources used is also widely varied. As in all countries, the media landscape in the US is not spared from political polarization, resulting in biases that are evident in almost all sources and originate from both sides of the political spectrum. Because of the wide variety of sources used, ACLED researchers are able to triangulate information gained from multiple sources, thus mitigating political bias. Furthermore, ACLED consistently reviews new sources of information to determine their ability to provide information on distinct events, and so is continuously expanding its scope of coverage. For more information on ACLED’s general sourcing methodology, see this sourcing primer.

Nationally, the make-up of ACLED’s US source list can be visualized using the Source Scale variable as well. Figure 2 below displays the US “sourcing profile”: an outline of the types of sources reviewed on a weekly basis. 

Because of the wide geographic dispersion and variety of media sources across the country, subnational sources (sub-regions, states, counties, municipalities), in teal, make up the vast majority of a researcher’s weekly review, referenced in nearly three-quarters of events. This includes sources such as the Oregonian, the Atlanta-Journal Constitution, St. Louis Post-Dispatch, and local TV stations affiliated with ABC, NBC, Fox, etc.

Other traditional media sources – categorized using the national (light blue); regional, or North American sources (in brown) , and international (in lightest blue) source scales – are referenced in fewer than three percent of events combined. These sources tend to capture only the most sensational events, and hence offer unique details to be coded in a limited number of cases.

New media sources, in gray, are also referenced to a limited degree: in less than 1 percent of events. ACLED does not crowdsource information as social media can be more easily susceptible to manipulation (e.g. “fake news”, bots, etc.). Rather, new media sources reviewed are trusted accounts, such as those of trusted freelance journalists or local state or police blogs.

Other sources, in orange, are referenced in nearly one-quarter of events. These include the US Press Freedom Tracker, or reports from political parties, labor unions, monitoring groups, such as Amnesty International, as well as demonstration data aggregators, such as the Count Love ² Coverage by the Count Love project came to an end in January 2021. project.³ While information from the Crowd Counting Consortium (CCC) was included in earlier coverage by ACLED, it is no longer included. Sources of information used by CCC which meet ACLED’s inclusion criteria, continue to be covered and are now integrated as part of ACLED’s source list directly. For the data aggregators, the original source of information (i.e. the name of the media outlet) is included as a source in addition to the name of the data aggregator project, in order to attribute credit to the fact that the data aggregator project directed ACLED to the original source. 

Data collected by local partners, in black, is a powerful supplemental source in ACLED’s coverage around the globe. In the case of the US, there was a noticeable deficit in coverage of militia activity outside of violent events and demonstrations in which they are directly involved. In order to capture and track non-violent activity of militias across the US, ACLED has partnered with MilitiaWatch, a research project and blog that tracks, documents, and analyzes contemporary US militia movements, and provides reports connecting long-term militia trends to broader political events. MilitiaWatch gathers data from open source and semi-open source content created by and for militia members, allowing for an analysis of militia activity from their own perspectives. MilitiaWatch partner data typically provide “Strategic development” events, such as recruitment drives, training exercises, the creation of new groups or splinter groups, or important announcements, such as a militia’s support for a political group. Furthermore, MilitiaWatch may provide more specific details as to the identity of militias active during demonstrations, allowing for the improvement of already published data. 

Beyond MilitiaWatch, ACLED is partnered with Network Contagion Research Institute (NCRI)⁴ This partnership began in June 2021., a not-for-profit that pushes for greater public safety protections via social media platforms and government bodies. NCRI shares information regarding the far-right and anti-COVID restriction movements in the US. Data from NCRI focuses on open source content regarding right-wing extremism. NCRI is a research coalition aimed at understanding how violent and hateful rhetoric can spread through online media environments. Previously, ACLED was partnered with Political Research Associates (PRA)⁵ This partnership ended in May 2021., a think tank focused on researching the movements and ideologies that undermine human rights. They shared information on the far-right in the US, especially with regards to antisemitism, white nationalism, and the Christian right, which came from secondary data through the monitoring of various online forums and media reports.

The supplementation of the data with new or more detailed information is an ongoing process. As a result, ACLED continues to pursue partnerships with other organizations or research projects whose data or expertise may improve the quality of the USA dataset. This section will thus be updated once new partnerships are established.