ACLED Codebook 2023
ACLED Codebook 2023
ACLED Codebook 2023
Table of Contents T
ACLED Codebook 1
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 2
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Introduction
ACLED collects reported information on the type, agents, location, date, and other
characteristics of political violence events, demonstration events, and other select
non-violent, politically-relevant developments in every country and territory in the
world. ACLED focuses on tracking a range of violent and non-violent actions by or
affecting political agents, including governments, rebels, militias, identity groups,
political parties, external forces, rioters, protesters, and civilians. The full list of ACLED
data columns is available in the table below.
• Tracking rebel, militia, and government activity over time and space;
• Recording violent acts between and across non-state groups, including
political and identity militias;
• Recording political violence by unnamed agents, as violent groups may
remain unnamed for strategic reasons;
• Recording attacks on civilians by all violent political agents;
• Distinguishing between territorial transfers of military control from
governments (and their affiliates) to non-state agents and vice versa;
• Collecting information on rioting and protesting; and
• Tracking non-violent strategic developments representing crucial junctures
in periods of political violence (e.g. recruitment drives, peace talks, high-level
arrests).
ACLED data are derived from a wide range of local, national, and international sources
in over 75 languages. The information is collected by trained researchers worldwide.
An updated overview of ACLED’s current country and time period coverage is available
here.
ACLED data are coded in real time and published on a weekly basis following a multi-
stage internal review process. Data can be downloaded through the export tool,
curated data files (which include regional data files and files for specific subsets like
violence targeting civilians), or API with a free account. Further guides to ACLED’s
methodology, coding decisions, and access system are available on the ACLED
website.
ACLED Codebook 3
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 4
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 5
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
region The region of the world where the • E.g. Eastern Africa
event took place.
ACLED Codebook 6
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 7
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Events
Political violence is defined as the use of force by a group with a political purpose
or motivation, or with distinct political effects.1 A political violence event is a single
altercation where force is used by one or more groups toward a political end. A
demonstration event is an in-person public gathering of three or more people
advocating for a shared cause. Other select non-violent instances of politically
significant developments are also included in the dataset to capture the potential
precursors or critical junctures of a violent conflict.
ACLED has developed a system that defines political disorder by its constituent events.
It begins with robust and broad definitions of political violence and demonstrations
(more below) and limits the taxonomy of political disorder to event types and sub-
event types. This taxonomy allows users to compare trends of political disorder
across countries and time periods. ACLED does not predefine broader aggregate
categories of events, like wars, conflicts, operations, campaigns, or movements.2
These analytical decisions are left to the user.
ACLED allows users to filter data based on the type of event, type of actor, type
of interaction, actors, location, or time period. ACLED does not dictate nor frame
aggregations of conflict events, but allows users to determine how to define and
explore frames. In this way, the dataset is designed to provide a comprehensive
overview of political disorder around the world.
1 ACLED includes organized criminal violence as political violence when it challenges safety, security,
and territorial control within a state. For more on this, see the ACLED Gang Methodology.
2 ACLED also does not categorize clusters or campaigns of events into categories of state or non-
state operations, terrorism, insurgency, etc. This is because political violence, demonstrations,
conflict, and disorder are aggregations of multiple actors, attacks, goals, etc. Classifying events into
parts of a civil war, livelihood, religious, or election conflict (to name a few common versions) is often
a reductive choice, leaving out the complex ways conflict creates multiple scales of violence and
intended outcomes within a territory.
ACLED Codebook 8
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
The fundamental unit of observation in ACLED is the event. Events involve designated
actors – e.g. a named rebel group, a militia, or state forces.3 They occur at a specific
named location (identified by name and geographic coordinates) and on a specific
day. Researchers work to ensure that the most specific possible location and time
are recorded. ACLED currently records six event types and 25 sub-event types, both
violent and non-violent. Sub-event types are also categorized by three overarching
disorder types to facilitate analysis: 1) ‘Political violence’, ‘Demonstrations’, and
‘Strategic developments’. Table 2 displays ACLED event types, sub-event types, and
disorder types.
Both event types and sub-event types are hierarchical to accommodate concurrent
tactics within the same event, to avoid double-counting (for more, see the section Event
Aggregation). This means that an airstrike (individually recorded as an ‘Explosions/
Remote violence’ event) occurring within the same context as a ground battle would be
recorded as one ‘Battles’ event. Likewise, an attack on a civilian (individually recorded
as a ‘Violence against civilians’ event) occurring within the same context as a remote
explosive detonation (individually recorded as an ‘Explosions/Remote violence’ event)
would be recorded as one ‘Explosions/Remote violence’ event. A similar structure
holds for sub-event types. Ground shelling occurring simultaneously as an airstrike
would be recorded as an ‘Air/drone strike’ event, as it is higher on the hierarchy than
the ‘Shelling/artillery/missile attack’ sub-event type. Likewise, a civilian abducted and
then killed would be recorded with the ‘Attack’ sub-event type because it is higher on
the hierarchy than the ‘Abduction/forced disappearance’ sub-event type. The event
types and sub-event types noted in Table 2 below are presented in hierarchical order.
It is important to keep these distinctions in mind when interpreting and drawing
conclusions from the data.
3 With the sole exception of Unidentified Armed Group and generic categories including Rioters,
Protesters, and Civilians.
ACLED Codebook 9
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 10
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Battles
ACLED defines a ‘Battles’ event as a violent interaction between two organized armed
groups at a particular time and location. ‘Battles’ can occur between armed and
organized state, non-state, and external groups, and in any combination therein. There
is no fatality minimum necessary for inclusion. Civilians can be harmed in the course of
larger ‘Battles’ events if they are caught in the crossfire, for example, or affected by
strikes on military targets, which is commonly referred to as ‘collateral damage’ (for
more, see Indirect Killing of Civilians). When civilians are harmed in a ‘Battles’ event,
they are not recorded as an ‘Associated Actor’, the event is not marked in the ‘Civilian
targeting’ column (for more, see Civilian targeting), nor is a separate civilian-specific
event recorded. If any civilian fatalities are reported as part of a battle, they are
aggregated in the ‘Fatalities’ column for the ‘Battles’ event, and they are described in
the ‘Notes’ column.
The ‘Battles’ event type may include: ground clashes between different armed groups,
ground clashes between armed groups supported by artillery fire or airstrikes,
ambushes of on-duty soldiers or armed militants, exchanges of artillery fire, ground
attacks against military or militant positions, air attacks where ground forces are able
to effectively fire on the aircraft, and air-to-air combat.
The following sub-event types are categorized under the ‘Battles’ event type and are
designated according to whether the battle does or does not result in a territorial
exchange: ‘Armed clash’, ‘Government regains territory’, and ‘Non-state actor
overtakes territory’.
ACLED Codebook 11
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
are not recorded as ‘Battles’ events. Instead, they are recorded as ‘Non-violent transfer
of territory’ under the ‘Strategic developments’ event type (more below).
This sub-event type is used when government forces or their affiliates that are
fighting against competing state forces or against a non-state group regain
control of a location through armed interaction. This sub-event type is only
recorded for the re-establishment of government control and not for cases
where competing non-state actors exchange control. Short-lived and/or small-
scale territorial exchanges that do not last for more than one day are recorded
as ‘Armed clash’. Possible ‘Interaction’ codes include: 11, 12, 13, 14, and 18.
This sub-event type is used when a non-state actor (excluding those operating
directly on behalf of the government) or a foreign state actor, through armed
interaction, captures territory from an opposing government or non-state
actor; as a result, they are regarded as having a monopoly of force within that
territory. Short-lived and/or small-scale territorial exchanges that do not last for
more than one day are recorded as ‘Armed clash’ events. In cases where non-
state forces fight with opposing actors in a location many times before gaining
control, only the final territorial acquisition is recorded as ‘Non-state actor
overtakes territory’. All other battles in that location are recorded as ‘Armed
clash’. Possible ‘Interaction’ codes include: 12, 13, 14, 18, 22, 23, 24, 28, 33, 34,
38, 44, 48, and 88.
Armed clash
This sub-event type is used when armed, organized groups engage in a battle,
and no reports indicate a significant change in territorial control (as described
above). Possible ‘Interaction’ codes include: 11, 12, 13, 14, 18, 22, 23, 24, 28, 33,
34, 38, 44, 48, and 88.
ACLED Codebook 12
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Protests
Protesters are noted by generic actor names: Protesters (Country). If they are
representing a group, the name of that group is recorded in the corresponding
‘Associated Actor’ column.
The following sub-event types are associated with the ‘Protests’ event type: ‘Peaceful
protest’, ‘Protest with intervention’, and ‘Excessive force against protesters’.
Excessive force against protesters
This sub-event type is used when individuals are engaged in a peaceful protest
and are targeted with lethal violence or violence resulting in serious injuries (e.g.
requiring hospitalization). This includes situations where remote explosives,
such as improvised explosive devices, are used to target protesters, as well
as situations where non-state actors, such as rebel groups, target protesters.
Possible ‘Interaction’ codes include: 16, 26, 36, 46, 56, and 68.
This sub-event type is used when individuals are engaged in a peaceful protest
during which there is a physical attempt to disperse or suppress the protest
without serious/lethal injuries or the targeting of protesters with lethal weapons
reported. This sub-event type also covers any instance where armed groups
or rioters interact with peaceful protesters without resulting in serious/lethal
ACLED Codebook 13
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Peaceful protest
This sub-event type is used when demonstrators gather for a protest and
do not engage in violence or other forms of rioting activity, such as property
destruction, and are not met with any sort of force or intervention. Possible
‘Interaction’ codes include: 60, 66, and 67.
Riots
‘Riots’ are violent events where demonstrators or mobs of three or more engage in
violent or destructive acts, including but not limited to physical fights, rock throwing,
property destruction, etc. They may engage individuals, property, businesses, other
rioting groups, or armed actors. Rioters are noted by generic actor names: Rioters
(Country). If rioters are affiliated with a specific group – which may or may not be
armed – or identity group, that group is recorded in the respective ‘Associated Actor’
column. Riots may begin as peaceful protests, or a mob may have the intention to
engage in violence from the outset. ‘Riots’ events where civilians are the main or only
target will be tagged with “Civilians targeted” in the ‘Civilian targeting’ column.
The following sub-event types are associated with the ‘Riots’ event type: ‘Violent
demonstration’ and ‘Mob violence’.
Violent demonstration
4 Dispersal orders or citations, in the absence of reports of physical action to detain or remove
protesters, are not captured under this sub-event type; if protesters are only issued citations, the
event would be coded as a ‘Peaceful protest’.
ACLED Codebook 14
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Mob violence
This sub-event type is used when rioters violently interact with other rioters,
civilians or their property, or armed groups outside of demonstration contexts.
A mob is considered a crowd of people that is disorderly and has the intention
to cause harm or disruption through violence or property destruction. Note
that this type of violence can also include spontaneous vigilante mobs clashing
with other armed groups or attacking civilians. While a ‘Mob violence’ event
often involves unarmed or crudely armed rioters, on rare occasions, it can
involve violence by people associated with organized groups and/or using more
sophisticated weapons, such as firearms.
Explosions/Remote violence
The following sub-event types are associated with the ‘Explosions/Remote violence’
event type: ‘Chemical weapon’, ‘Air/drone strike’, ‘Suicide bomb’, ‘Shelling/artillery/
missile attack’, ‘Remote explosive/landmine/IED’, and ‘Grenade’.
Chemical weapon
ACLED Codebook 15
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Air/drone strike
This sub-event type is used when air or drone strikes take place in the absence
of any other engagement. Please note that any air-to-ground attacks fall
under this sub-event type, including attacks by helicopters that do not involve
exchanges of fire with forces on the ground.
Suicide bomb
This sub-event type is used when a suicide bombing occurs in the absence
of an armed clash, such as an exchange of small arms fire with other armed
groups. It also includes suicide vehicle-borne improvised explosive device
(SVBIED) attacks. Note that the suicide bomber is included in the total number
of reported fatalities coded for such events.
Shelling/artillery/missile attack
This sub-event type captures the use of long-range artillery, missile systems,
or other heavy weapons platforms in the absence of any other engagement.
When two armed groups exchange long-range fire, it is recorded as an ‘Armed
clash’. ‘Shelling/artillery/missile attack’ events include attacks described as
shelling, the use of artillery and cannons, mortars, guided missiles, rockets,
grenade launchers, and other heavy weapons platforms. Crewed aircraft shot
down by long-range systems fall under this sub-event type. Uncrewed armed
drones that are shot down, however, are recorded as interceptions under
‘Disrupted weapons use’ because people are not targeted (see below). Similarly,
an interception of a missile strike itself (such as by the Iron Dome in Israel) is
also recorded as ‘Disrupted weapons use’.
Remote explosive/landmine/IED
ACLED Codebook 16
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
SVBIEDs are recorded as ‘Suicide bomb’ events (see above), while the safe
defusal of an explosive or its accidental detonation by the actor who planted it
(with no other casualties reported) is recorded under ‘Disrupted weapons use’
(see below).
Grenade
This sub-event type captures the use of a grenade or any other similarly hand-
thrown explosive, such as an IED that is thrown, in the absence of any other
engagement. Events involving so-called ‘crude bombs’ (such as Molotov cocktails,
firecrackers, cherry bombs, petrol bombs, etc.) as well as ‘stun grenades’ are
not recorded in this category, but are included under either ‘Riots’ or ‘Strategic
developments’ depending on the context in which they occurred.
In cases where the identity and actions of the targets are in question (e.g. the target
may be employed as a police officer), ACLED determines that if a person is harmed
or killed while unarmed and unable to either act defensively or counter-attack, this
is an act of ‘Violence against civilians’. This includes extrajudicial killings of detained
combatants or unarmed prisoners of war.
‘Violence against civilians’ also includes attempts at inflicting harm (e.g. beating,
shooting, torture, rape, mutilation, etc.) or forcibly disappearing (e.g. kidnapping and
disappearances) civilian actors. Note that the ‘Violence against civilians’ event type
ACLED Codebook 17
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
exclusively captures violence targeting civilians that does not occur concurrently with
other forms of violence – such as rioting – that are coded higher in the ACLED event
type hierarchy. To get a full list of events in the ACLED dataset where civilians were
the main or only target of violence, users can filter on the ‘Civilian targeting’ column
(for more, see Civilian targeting).
The following sub-event types are associated with the ‘Violence against civilians’ event
type: ‘Sexual violence’, ‘Attack’, and ‘Abduction/forced disappearance’.
Sexual violence
This sub-event type is used when any individual is targeted with sexual violence.
‘Sexual violence’ is defined largely as an action that inflicts harm of a sexual
nature. This means that it is not limited to solely penetrative rape, but also
includes actions like public stripping, sexual torture, etc. Given the gendered
nature of sexual violence, the gender identities of the victims – i.e. Women
(Country), Men (Country), and LGBTQ+ (Country), or a combination thereof –
are recorded in the ‘Associated Actor’ column for these events when reported.
Note that it is possible for sexual violence to occur within other event types
such as ‘Battles’ and ‘Riots’ (for more, see Sexual Violence in the ACLED Dataset).
Attack
This sub-event type is used when civilians are targeted with violence by an
organized armed actor outside the context of other forms of violence coded
higher in the ACLED event type hierarchy. Attacks of a sexual nature are recorded
as ‘Sexual violence’ (see above). Violence by law enforcement that constitutes
severe government overreach is also recorded as an ‘Attack’ event.
Abduction/forced disappearance
This sub-event type is used when an actor engages in the abduction or forced
disappearance of civilians, without reports of further violence. If fatalities or
serious injuries are reported during the abduction or forced disappearance,
the event is recorded as an ‘Attack’ event instead. If such violence is reported
in later periods during captivity, this is recorded as an additional ‘Attack’ event.
Note that multiple people can be abducted in a single ‘Abduction/forced
disappearance’ event.
ACLED Codebook 18
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Strategic developments
This event type captures contextually important information regarding incidents and
activities of groups that are not recorded as ‘Political violence’ or ‘Demonstrations’
events, yet may trigger future events or contribute to political dynamics within and
across states. The inclusion of such events is limited, as their purpose is to capture
pivotal events within the broader political landscape. They typically include a disparate
range of events, such as recruitment drives, looting, and incursions, as well as the
location and date of peace talks and the arrests of high-ranking officials or large
groups. While it is rare for fatalities to be reported as a result of such events, they
can occur in certain cases – e.g. the suspicious death of a high-ranking official, the
accidental detonation of a bomb resulting in the bomber being killed, etc.
Due to their context-specific nature, ‘Strategic developments’ are not collected and
recorded in the same cross-comparable fashion as ‘Political violence’ and ‘Demonstration’
events. As such, the ‘Strategic developments’ event type is primarily a tool for
understanding particular contexts (see this primer on using ‘Strategic developments’).
The following sub-event types are associated with the ‘Strategic developments’
event type: ‘Agreement’, ‘Arrests’, ‘Change to group/activity’, ‘Disrupted weapons
use’, ‘Headquarters or base established’, ‘Looting/property destruction’, ‘Non-violent
transfer of territory’, and ‘Other’.
Agreement
This sub-event type is used to record any sort of agreement between different
actors (such as governments and rebel groups). Examples include peace
agreements/talks, ceasefires, evacuation deals, prisoner exchanges, negotiated
territorial transfers, prisoner releases, surrenders, repatriations, etc.
5 In rare cases where non-state groups are able to maintain some level of judicial/penal system, they
are also considered able to engage in ‘Arrests’, and these actions are not recorded as ‘Abduction/
forced disappearance’ and are instead recorded under the ‘Arrests’ sub-event type.
ACLED Codebook 19
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Arrests
This sub-event type is used when state forces or other actors exercising de
facto control over a territory either detain a particularly significant individual or
engage in politically significant mass arrests.
Change to group/activity
Looting/property destruction
This sub-event type is used when actors engage in looting or seizing goods or
ACLED Codebook 20
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Other
This sub-event type is used to cover any significant development that does not
fall into any of the other ‘Strategic developments’ sub-event types. Examples
include the occurrence of a coup, the displacement of a civilian population as a
result of fighting, or the discovery of mass graves.
Event Aggregation
One or more events can occur in the same location on the same day. If two different
types of interactions between the same actors in the same location on the same day
are reported, they are typically noted as a single aggregate event. For example:
1. A rebel group fights with government forces in a town and wins control.
Rebel artillery strikes are reported throughout the day. In this case, only
a single ‘Battles’ event between the rebels and the government forces is
recorded, instead of one ‘Battles’ event and another ‘Explosions/Remote
violence’ event.
2. On the same day, demonstrators hold peaceful protests that progress into
violent clashes with security forces. In this case, a single ‘Riots’ event is
recorded, instead of one ‘Protests’ event and another ‘Riots’ event.
ACLED takes steps to avoid double-counting incidents, which can entail the aggre-
gation of different forms of violence into single events. For example, if civilians are
ACLED Codebook 21
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
killed in the context of a larger ‘Battles’ event, then these reported fatalities will be
added to the total number of fatalities reported for the ‘Battles’ event. The killing of
civilians will be noted in the ‘Notes’ column of the ‘Battles’ event, but it will not consti-
tute a separate event. In ‘Battles’ events, civilians are never recorded as ‘Associated
Actors’ as both parties are armed actors assumed to engage in violence.
Further, if an assault – such as an airstrike – hits a military target, but reports also
indicate that civilians were harmed, civilians are recorded as ‘Associated Actors’. Sim-
ilar to the previous case, the fatalities from this event, if reported, are aggregated
together. In cases where a bombing occurs with a vague and unspecified military
target (e.g. a bombing in a city where some militants are present, rather than an area
known to be controlled and actively used by militants), but civilians are the main
group affected (e.g. civilians are injured or killed), they will be recorded in the ‘Actor
2’ column. Militants may appear in the ‘Associated Actor 1’ and ‘Associated Actor 2’
columns.
However, if another event type involving different actors occurs on the same day and
in the same location, it is recorded separately. Hence, it is possible to have multiple
events – involving distinct actors – occur in the same location on the same day. For
example, if an incident of armed violence separately targets civilians on the same day
and location of a battle (e.g. a group of militants abducts civilians after engaging po-
lice in a shootout), two events are recorded to accurately capture the ‘Battles’ event
and the distinct ‘Violence against civilians’ event.
In most cases, an event requires two actors, noted in columns ‘Actor 1’ and ‘Actor
2’. However, the ‘Explosions/Remote violence’, ‘Riots’, ‘Protests’, and ‘Strategic devel-
opments’ event types can include one-sided events. If more than two actors are re-
ported, only the most important engagement is recorded, and the additional groups
may be recorded as ‘Associated Actors’ depending on the context (e.g. police forces
intervening to disperse rioters). The order of actors has no meaning in the ACLED data-
set, barring cases of ‘Violence against civilians’, in which the target is always recorded
as ‘Actor 2’.
Civilian targeting
In order to facilitate the analysis of all events in the ACLED dataset that feature vio-
lence targeting civilians, the ‘Civilian targeting’ column allows for filtering of events in
ACLED Codebook 22
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
which civilians were the main or only target of an event. Besides events coded under
the ‘Violence against civilians’ event type, civilians may also be the main or only target
of violence in events coded under the ‘Explosions/Remote violence’ event type (e.g.
a landmine killing a farmer), ‘Riots’ event type (e.g. a village mob assaulting another
villager over a land dispute), and ‘Excessive force against protesters’ sub-event type
(e.g. state forces using lethal force to disperse peaceful protesters). Events in which
civilians were incidentally harmed are not included in this category.
Actors
ACLED codes a range of actors, including state forces, rebels, militias, identity groups,
demonstrators, civilians, and external or other forces.
The vast majority of political violence events involve organized armed actors, such
as government forces and their affiliates, rebel groups, militias, external or private
forces (e.g. United Nations missions), and other political groups that interact over
issues of political authority (e.g. territorial control, government control, access to re-
sources, etc.). All organized armed actors are named,6 have a political purpose, and
use violence for political means. In order for an actor to be classified as an organized
agent of political violence – as opposed to a disorderly riotous group – the organi-
zation must be cohesive and assembled for more than a single event. Further, the
activity of such an organization must be connected to a means for achieving a larger
political purpose. This necessary and sufficient definition of organized actors allows
for the establishment of campaigns and trajectories of movements. In contrast to
armed organized actors, rioters are considered to be unorganized in the sense that
they lack inherent organization beyond spontaneous or atomic acts of organization.
Protesters and civilians may be organized and will have named organizations record-
ed where relevant.
The name of each actor is noted in the ‘Actor 1’ and ‘Actor 2’ columns of the dataset.
The group type is recorded in the ‘Inter 1’ and ‘Inter 2’ columns, while their dual en-
gagement is noted in the ‘Interaction’ column.
The ‘Associated Actor 1’ and ‘Associated Actor 2’ columns record the associated
6 Except for the actors marked with “unidentified”, e.g. Unidentified Armed Group or Unidentified
Gang.
ACLED Codebook 23
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
groups for specific events as well as the identity of specific actors. In the former case,
an associated group may be an ally in the action – such as two organized armed
groups that are engaging in attacks against a common enemy. In the latter case, the
‘Associated Actor 1’ and ‘Associated Actor 2’ columns may record additional informa-
tion concerning the targets of an attack, socio-political affiliation of demonstrators,
or ethno-religious identity of a civilian victim. An ‘Explosions/Remote violence’ event
that destroys the base of an organized armed group, but also affects civilians, will
have both groups noted – the primary actor will be the ostensibly-intended target
(i.e. the armed group in this case), while civilians (commonly described as ‘collateral
damage’ in this case) would be the ‘Associated Actor’.7
ACLED records the recognized name of actors as reported, whenever possible. In ex-
ceptional circumstances described in detail below, the name of an actor is generated
to reflect their origins and composition.
Each named actor is also designated as a type of organization. There are thousands
of individually named actors within the ACLED dataset, and the ‘Inter’ code categoriz-
es actors by whether they have similar organizational structures, goals, and practic-
es. Group type designates all actors into one of eight ACLED categories and assigns a
number in the ‘Inter 1’ and ‘Inter 2’ columns to that actor.
These categories offer a way to distinguish between actors and determine how pat-
terns of activity conform to goals and organizations. ACLED does not use a pattern
of activity to designate what kind of agent an actor is. ACLED specifically observes the
goals and structure of an actor and, where possible, its spatial dimension and relation-
ships to communities.
As such, the ‘Inter’ code of an actor can change over time. For example, if a rebel
group is successful in overthrowing a regime or seceding from a state, its armed
agents may then become the armed wing of a political party within the new regime
7 While civilians are recorded as an ‘Associated Actor’ when they are incidentally killed in ‘Explosions/
Remote violence’ attacks, they are not recorded as such in a ‘Battles’ event. This is because events
under the former involve one side not engaging in any violence within the course of the event; in a
‘Battles’ event, both sides, by definition, are capable of engaging in or responding with violence, and it
would hence be incorrect to denote that civilians are associated with either side.
ACLED Codebook 24
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
structure – effectively, a change in ‘Inter’ code from 2 to 3, e.g. the Imbonerakure mi-
litia in Burundi – or the government forces of the new state – a change in ‘Inter’ code
from 2 to 1, e.g. the Taliban turned into Afghanistan’s state forces.
Certain types of violent agents may appear to fall outside of this categorization, but
ACLED has designed these classifications with the flexibility to fit the universe of
agents operating in conflict. For example, militant religious organizations can have
various goals (e.g. al-Qaeda), including overtaking the state, influencing political pro-
cesses and supporting regional political elites, and engaging in communal contests
over access to religious sites. ACLED does not allow for “insurgents” or “terrorists” as
types of agents, but instead categorizes actors as rebels, militias, communal organiza-
tions, protesters, etc. Many armed organizations may use insurgency tactics or com-
mit acts of violence against civilians, intending to cause a large number of fatalities
as part of their violent repertoire. Instead, ACLED considers the goal and organization of
each group to be the basis for their classification.
State forces are defined as collective actors, including military and police,
that are recognized to perform government functions over a given territory.
Government actors are named by ACLED as a series of separate regimes rather
than a uniform body. For example, the government regimes in the present-
day Democratic Republic of Congo since 1965 are recorded as Congo/Zaire
(1965-1997), Democratic Republic of Congo (1997-2001), and Democratic
Republic of Congo (2001-2019) as opposed to Congo/Zaire (1965-present). As
the strength, capacity, and policies of governments can vary widely from one
regime to the next, ACLED designates governments by their leading regimes.
This enables researchers to capture the differences in government involvement
and reaction to violence.
As militaries and police forces are a direct arm of the government, these actors
are noted as Military Forces of State (Year-Year) or Police Forces of State
(Year-Year). Mutinies of militaries are recorded as Mutiny of Military Forces
of State (Year-Year). Various units of these state forces are recorded distinctly
as well – such as Police Forces of India (2014-) Assam Rifles or Police Forces
of the Philippines (2016-) Anti-Illegal Drugs Special Operations Task
Force – given that such units can engage in distinct patterns of behavior. Pro-
government militias with indirect links to the state are not included as state
ACLED Codebook 25
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
forces, given their deliberate distance from formal ties to the state.
It is important to note that this classification of state forces does not imply
legitimacy, but rather acknowledges the de facto exercise of authority over a
territory. This is why, in a limited number of cases, ACLED records government
actors in states with limited or no recognition as state forces. These states are not
necessarily recorded separately in the ‘Country’ column, but their government
forces are recorded as distinct state actors. For example, the Military Forces of
Somaliland (1991-) actor is recorded, though its country of activity is recorded
in the ‘Country’ column as Somalia, not Somaliland.
Rebel groups are defined as political organizations with the goal of countering
an established national governing regime through violence. Rebel groups are
named according to the title they publicly use to represent themselves. The
designation as a rebel group means that the group has a stated political agenda
for national power (either through regime replacement or separatism), is
acknowledged beyond the ranks of immediate members, and uses violence as
its primary means to pursue political goals.
Rebel forces are known by a specific chosen name and the groups are open and
transparent about their intentions and leadership. They typically operate within
and across states, and conduct activity against the central governments and
their associates. Rebel groups often have predecessors and successors due to
diverging goals within their membership. If splinter groups or factions within a
group emerge, these are recorded as distinct actors.
In cases where aggregate groups are contesting the government, we often use
an overarching name rather than factions (e.g. Hutu Rebels active in Burundi;
ACLED Codebook 26
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Political militias are a diverse set of violent actors that are often created for a
specific purpose or during a specific time period (e.g. the Janjaweed, largely
active in Sudan) and for the furtherance of a political goal by violence. Political
militias are recorded by their stated name. These organizations are defined
by their political goals of influencing and impacting governance, security, and
policy. However, these groups are not seeking the removal of a national power,
but are instead typically supported, armed by, or allied with a political elite
and act towards a goal defined by these elites or larger political movements.
Political militias often operate in conjunction, or in alliance, with a recognized
government, governor, military leader, rebel organization, political party,
business elite, or opposition group. Whereas some opposition parties have
a militia arm, groups such as the Sudanese Janjaweed or Serbian Tigers are
pro-government militias8 that work as supplements to government power yet
maintain indirect links to such power.
8 Clionadh Raleigh and Roudabeh Kishi, ‘Hired Guns; Using Pro-Government Militias for Political
Competition,’ Terrorism and Political Violence, 8 January 2018
ACLED Codebook 27
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
remain unknown to the larger public in order to pursue violence with impunity.
ACLED records such unidentified actors using the name Unidentified Armed
Group (Country).9
ACLED includes a broad category of identity militias for armed and violent
groups organized around a collective, common feature, including: community,
ethnicity, region, religion, or – in exceptional cases – livelihood. Therefore, for
ACLED’s purposes, identity militias include those identified as tribal, communal,
ethnic, local, clan, religious, and caste militias in source reporting. Events
involving identity militias are often referred to as communal violence, as these
violent groups typically act locally in pursuit of local goals, resources, power,
security, and retribution.
9 An Unidentified Armed Group is recorded with an ‘Inter’ code of 3, unless they are Unidentified
Military Forces (a rare case that has an ’Inter’ code of 1 or 8’ depending on the context) or Unidentified
Ethnic/Communal/Clan/Tribal Militia with an ‘Inter’ code of 4, as discussed below.
ACLED Codebook 28
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Interaction Codes
The joined ‘Interaction’ code is the combination of the two ‘Inter’ codes associated
with the two main actors in an event. Single actor type codes are recorded in ‘Inter 1’
10 Though non-violent, protesters may engage in disruptive behavior like burning effigies or shoe-
throwing.
ACLED Codebook 29
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
and ‘Inter 2’ columns, and the compounded number is recorded in the ‘Interaction’
column. For example, if a country’s military fights a political militia group, and the
respective ‘Inter 1’ and ‘Inter 2’ codes are 1 and 3, respectively, the compounded
‘Interaction’ code is recorded as 13.
‘Interaction’ code numbers are always the smallest possible number (for example,
37 instead of 73), regardless of the order of ‘Actor 1’ and ‘Actor 2’. ‘Interaction’ codes
are recorded for all events, including non-violent activity. For one-sided events, the
empty ‘Actor 2’ column is recorded as 0. For example, if an event involving a rebel
group is non-violent, the ‘Inter 1’ code is 2, and 20 is recorded in the ‘Interaction’
column. Only the actors recorded in the ‘Actor 1’ and ‘Actor 2’ columns form the basis
for the ‘Interaction’ code of an event, and not the actors recorded in the ‘Associated
Actors’ columns.
ACLED Codebook 30
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 31
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 32
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 33
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
ACLED Codebook 34
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Event Geography
There are up to six different types of spatial information that can be recorded for
each ACLED event:
1. The continental sub-region in which the event occurred;
2. The country in which the event occurred and its associated ISO code;
3. The name of the first, second, and third level administrative zones in which
the specific location is found, according to GIS-based assessments and
updated administrative codes;
4. The name of the specific location of an event;
5. The geographic coordinates of that specific location; and
6. A spatial precision code.
The most specific location for an event is sought for each event ACLED records, using
multiple sources to triangulate better location information.
In selected large cities with activity dispersed over many neighborhoods, locations
are further specified to predefined subsections within a city to prevent excessive
aggregation of events to a single city location. In such cases, locations are recorded
as: City Name - District name (e.g. Mosul - Old City) in the ‘Location’ column. If
information about the specific neighborhood/district is not known, the location is
recorded at the city level (e.g. Mosul). In both cases, ‘Geo-precision’ code 1 is recorded
(see the section below on spatial precision codes). The hyphenation feature allows users
to aggregate events by city, if needed.
If the source reporting indicates a particular town, and coordinates are available for
that town, the highest precision level, ‘Geo-precision’ code 1, is recorded. If the source
material indicates that activity took place in a small part of a region, and mentions
a general area, the event is coded to a town with geo-referenced coordinates to
ACLED Codebook 35
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
represent that area, and the ‘Geo-precision’ code 2 is recorded. If activity occurs near
a town or a city, this same ‘Geo-precision’ code 2 is employed. If a larger region is
mentioned, the closest natural location noted in reporting (like “border area,” “forest,”
or “sea,” among others) – or a provincial capital is used if no other information at all is
available – is chosen to represent the region, and ‘Geo-precision’ code 3 is recorded.
Barring some ‘Strategic developments’ events, ACLED does not record events that are
imprecise to the country level – meaning reports that mention 1,000 conflict fatalities
“across the country” will be investigated until they can be recorded as a set of more
specific events and will not be recorded as a single countrywide event.
Event time
Dates are a necessary component of all ACLED events. ACLED events are atomic
as events are recorded by day. If a military campaign in an area starts on 1 March
2020 and lasts until 5 March 2020 with violent activity reported on each day, this is
recorded as five different events, with a different date for each entry. This episode
would not be entered as a single campaign of violence. This allows ACLED to record
the exact number of active days of a conflict. Events that the source material indicates
as occurring in the space of three months – such as long-running protests – are not
automatically assumed to have 90 days of activity, but instead are only recorded for
the days for which it is specifically reported that an activity took place. This avoids
over-counting event occurrences.
If the source material includes an actual date of an event, a ‘Time precision’ code
of 1 is recorded. If the source material indicates that an event happened sometime
during the week or within a similar period of time (i.e. during the weekend), a ‘Time
precision’ code of 2 is recorded and the middle of the period is used as the reference
date (i.e. with an event occurring last week, the ‘Event date’ would be recorded on
either Wednesday or Thursday of that week). If the source material only indicates
that an event took place sometime during a month (i.e. in the past two or three
ACLED Codebook 36
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
weeks, or in January), without reference to the particular date, the month mid-point
is chosen. If the beginning or end of the month is noted, the first and last date is
used, respectively. In both of these cases, a ‘Time precision’ code of 3 is recorded.
ACLED does not include events with less temporal information.
Reported Fatalities
Events that result in fatalities have the relevant information recorded in the
‘Fatalities’ column. Events recorded by ACLED do not have to meet a minimum fatality
number for inclusion. ACLED only records estimated fatalities when reported by
the source material. When and where possible, ACLED researchers seek out
information to triangulate the fatality numbers reported by sources, but ACLED does
not independently verify reported fatality estimates. Additionally, ACLED is a living
dataset, so all fatality figures are revised and corrected — upward or downward — if
new or better information becomes available (which, in some cases, can be months
or years after an event has taken place). Fatality data are typically the most biased,
and least accurate, component of conflict reporting. They are particularly prone to
manipulation by armed groups, and occasionally the media, which may overstate
or underreport fatalities for political purposes. These figures should therefore be
understood as indicative estimates of reported fatalities, rather than definitive fatality
counts.
If the source material differs or a vague estimate is provided, the lowest number
of reported fatalities is recorded in the ‘Fatalities’ column until a more reliable or
corroborated estimate becomes available. If reports mention “several,” “many,” “few,”
or plural “fatalities,” yet the exact number is unknown, either 3 or 10 is recorded as
the total, depending on the scale and context of the attack (for more, see FAQs: ACLED
Fatality Methodology). If a report mentions “dozens,” this is recorded as 12 reported
fatalities. If a report mentions “hundreds,” this is recorded as 100 reported fatalities.
If there is no reference made to fatalities in the source material, or if it is unclear
whether fatalities occurred at all (for example, when only “casualties” are mentioned,
which – by definition – means ‘injuries and/or fatalities’), 0 reported fatalities are
recorded.
When summarized fatalities are reported, but events occur across several days or
in multiple locations simultaneously (e.g. “12 fatalities result from fighting over a
span of 3 days”), the total number of reported fatalities is divided and that fraction is
ACLED Codebook 37
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
recorded for each day of the event (4 reported fatalities per battle day, in the example
above). If an odd number of fatalities is recorded (including 1), the proportion of
reported fatalities is divided by assigning the first day the additional fatality and it is
then distributed as evenly as possible. The disaggregation of fatalities is recorded in
the ’Notes’ column of the relevant events.
Notes
The ‘Notes’ column summarizes the main features of the event. Notes are kept short
and limited to significant details about the specific event. In rare cases, additional
relevant information is added to provide context to the event.
Tags
The ‘Tags’ column provides additional structured information about an ACLED event.
Tags provide a flexible means of refining or grouping existing variables, or providing
new variables for specific contexts that are not already captured within the other data
columns. Tags can be specific to certain actors and event types, themes, countries, time
periods, etc., or combinations thereof. As such, some tags are hyperspecific, such as
the “count every vote” tag that was introduced for the 2020 election demonstrations
in the United States. Other tags, like the “crowd size” tag, are applied globally to all
‘Protests’ and ‘Riots’ events. Tags can be introduced or retired as situations on the
ground and analytical needs develop or subside.
Each tag is separated by a semicolon (;) in the ‘Tags’ column. Users may filter by
the tags to quickly generate subsets of the data on specific political violence or
demonstration activity (for more on how tags are added to the ACLED dataset and the
full list of tags, see Tags in ACLED Data).
ACLED Codebook 38
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
3. Local Partner Data: The past decades have seen an increase in organizations
established at the local level as both social activism and the ability to report
political violence have increased. These organizations leverage their local
knowledge as they collect and obtain information through primary and/
or secondary means. ACLED develops relationships with local partners to
enhance the depth and quality of its data.11
4. New Media (targeted and verified): ‘New media’ (e.g. Twitter, Telegram,
WhatsApp) can be a powerful supplemental source, but varies widely in terms
of quality. Therefore, ACLED does not crowdsource or scrape large amounts of
social media. Rather, a targeted approach to the inclusion of new media is
preferred through either the establishment of relationships with the source
directly, or the verification of the quality of each source.
Every ACLED event is based on at least one source. The source names or acronyms
are noted in the ‘Source’ column. With the exception of certain local sources that
wish to remain anonymous, the ‘Source’ column details are sufficient to retrace the
sources that have been used to record an event. All sources listed have contributed
information to the event. Researchers often find multiple reports confirming details
about an event; when multiple sources report on the same information, the most
thorough, reliable, and recent report is cited.
If that event has additional information taken from more than one source, both
report sources are noted in the ‘Source’ column.
11 For safety reasons, some partners that wish to remain anonymous are recorded as ‘Undisclosed
Source’ in the ‘Source’ column.
ACLED Codebook 39
THE ARMED CONFLICT LOCATION & EVENT DATA PROJECT
www.acleddata.com | admin@acleddata.com
Source scale
The ‘Source scale’ column describes whether the sources used for coding an event
operate at the local, subnational, national, regional, international, or other level.
ACLED develops tailored sourcing profiles for different conflict and disorder contexts
drawing on sources across various scales. An integral part of this sourcing strategy is
ACLED’s unique network of local data collection partners and observatories around
the world. With this network, ACLED is able to gather information in hard-to-access
contexts and integrate additional thematic, regional, and country-level expertise into
the dataset (for more, see FAQS: ACLED Sourcing Methodology).
As ACLED records a wide range of constituent events for periods and areas affected
by political violence and disorder, more information is collected than datasets that
focus on specific forms of violence (e.g. civil war, ‘terrorism’ as defined by targeting of
civilians by non-state actors alone, or those with fatality thresholds for inclusion) or
campaigns of violence rather than events. Please note that ACLED is a living dataset
and event details are subject to change as new or better information becomes
available. Results of analysis based on ACLED data may differ from analysis based on
other datasets with different scopes and methodologies.
ACLED Codebook 40