What does ACLED data cover?

ACLED collects realtime and historical data on political violence and protest in Africa from 1997 to the present, and in Asia from 2010 to the present (Asia coding underway). Data are published online, both in annual dataset format, and in realtime updates.

 


How are the data disaggregated?

The ACLED dataset seeks to provide information about political violence and protest in Africa and Asia which is disaggregated by date (when the event happened); type of violence (what happened); actors (who is involved); and location (where the event happened). Reports of violence are broken down into individual, discrete events, determined by whether they took place at a different time, involved different types of violence or actors, or occurred in different locations. In practical terms, this means that events that take place on different days, involving different types of violence, with different types of actors or in a different location are all coded as separate events.

Additional information on ACLED methodology can be found online in our codebook and user guide.

 


How are the data collected?

ACLED data are coded by a range of experienced researchers who collect information primarily from secondary sources and code it according to the coding outlined in the codebook.

ACLED takes a variety of steps to help ensure that the data we publish are both accurate as well as accessible for academics, policymakers, and practitioners to work with and integrate into their own research. ACLED data are collected each week after individual coders have scrutinized information from available reports; they are then aggregated and revised by the first coding reviewer, investigated and cross-checked by the second reviewer and then event notes and details are inspected by the third and final reviewer.

In addition to weekly review of the real-time conflict data, an overarching cleaning and data management effort is undertaken on an annual basis, leading to the annually-released dataset publication.

For additional information about data collection and review, users can consult the methodology page on our website and our working paper on reporting sources.

 


Can I use ACLED for my own research?

Yes, ACLED data are publicly available without charge. The data included in it is not confidential and so can be shared and analysed publicly. ACLED data can be used for academic and policy research, as well as analysis specific to organisations operating in target countries. All users are asked to cite Raleigh, Clionadh, Andrew Linke, Håvard Hegre and Joakim Karlsen. 2010. Introducing ACLED-Armed Conflict Location and Event Data. Journal of Peace Research 47(5) 651-660.

 


How do I cite ACLED data?

If using ACLED data in a written report, article or book, please kindly cite: Raleigh, Clionadh, Andrew Linke, Håvard Hegre and Joakim Karlsen. 2010. Introducing ACLED – Armed Conflict Location and Event Data. Journal of Peace Research 47(5), 651-660.

If referring to figures or statistics published in ACLED Conflict Trends reports, regional profiles, working papers, etc., please kindly cite the individual report, website and page number, for example: Armed Conflict Location & Event Data Project, Conflict Trends Report, No. 24, April 2014, p. 3, http://www.acleddata.com/research-and-publications/conflict-trends-reports/.

If using ACLED data in a visual, graphic, map or infographic, please kindly attribute the source data on the visual itself or within the key / legend. Examples can be found at:

If you wish to reproduce or publish an image, graph or map we have already published (rather than creating an original image using raw data), please advise of the image and its location (publication name, date and hyperlink if available). We review requests for reproduction on a case-by-case basis.

 


What do the specific columns mean in ACLED data?

In the ACLED dataset, each column contains specific information on the exact location, date, actors and other characteristics of politically violent events. Information in the dataset includes:

  • GWNO: A numeric code for each individual country from Gleditsch and Ward (2009).
  • EVENT_ID_CNTY: An individual identifier by number and country acronym.
  • EVENT_ID_NO_CNTY: An individual numeric identifier.
  • EVENT_DATE: Recorded as Day / Month / Year.
  • YEAR: The year in which an event took place.
  • TIME_PRECISION: A numeric code indicating the level of certainty of the date coded for the event.
  • EVENT_TYPE: The type of conflict event (please see below).
  • ACTOR1: The named actor involved in the event. If a dyadic event, this will be accompanied by a second actor. If a monadic event, no second actor necessary.
  • ALLY_ACTOR_1: The named actor allied with or identifying ACTOR1 in one specific event.
  • INTER1: A numeric code indicating the type of ACTOR1.
  • ACTOR2: The named actor involved in the event. If a dyadic event, there will be an “Actor 1”.
  • ALLY_ACTOR_2: The named actor allied with or identifying ACTOR2 in one specific event.
  • INTER2: A numeric code indicating the type of ACTOR2.
  • INTERACTION: A numeric code indicating the interaction between types of ACTOR1 and ACTOR2. Coded as an interaction between actor types, and recorded as lowest joint number.
  • COUNTRY: The country in which the event took place.
  • ADMIN1: The largest sub-national administrative region in which the event took place.
  • ADMIN2: The second largest sub-national administrative region in which the event took place.
  • ADMIN3: The third largest sub-national administrative region in which the event took place.
  • LOCATION: The location in which the event took place.
  • LATITUDE: The latitude of the location.
  • LONGITUDE: The longitude of the location.
  • GEO_PRECISION: A numeric code indicating the level of certainty of the location coded for the event.
  • SOURCE: The source of the event report.
  • NOTES: A short description of the event.
  • FATALITIES: Number or estimate of fatalities due to event. These are frequently different across reports.

Additional information on the ACLED dataset can be found online in our codebook and user guide.

 


What is the difference between event types?

ACLED currently codes eight types of events, both violent and non-violent, that may occur during a conflict. These include:

  • Battle-No change of territory (a battle between two violent armed groups where control of the contested location does not change);
  • Battle-Non-state actor overtakes territory (a battle where non-state actors win control of location);
  • Battle-Government regains territory (a battle in which the government regains control of a location);
  • Headquarters or base established (a non-state group establishes a base or headquarters without using violence);
  • Strategic development (accounts for often non-violent activity by conflict and other agents within the context of the war/dispute. Recruitment, looting and arrests are included);
  • Riots/Protests (violent and non-violent demonstrations, often by spontanous groups of civilians and against a government institution);
  • Violence against civilians (attacks by violent groups on civilians. No fatalities are necessary for inclusion);
  • Non-violent transfer of territory (situations where rebels or governments acquire control of a location without engaging in a violent act);
  • Remote violence (events involving bombings or similar attacks from a remote location, not requiring the physical presence of the perpetrator).

Additional information on event types in ACLED can be found online in our codebook and user guide. For additional information about data collection and review, users can consult the methodology page on our website and our working paper on reporting sources.

 


Why does ACLED release multiple data file formats, and how do I use these?

ACLED makes its dataset of disaggregated conflict data publicly available. A new version of the dataset is released annually, with data from the previous year and targeted quality review being added in each new version. Version 7 of the ACLED dataset covers political violence and protest in African states from January 1997 – December 2016 and is available online here. In addition to the published datasets which cover full years, ACLED produces up-to-date data on political violence and protest in Africa in 2017 every week. The data are published every Monday afternoon and is available on the realtime data page.

Users can also recreate their own dataset merging Version 6 of the ACLED dataset with realtime data, should they wish to include the most recent data in their analysis. To do this, users can copy and paste Version 6 and realtime data into a new worksheet.

ACLED datasets are available to the public in three formats: a Microsoft Excel sheet (or csv) containing data on all coded events which occur in states or continents; a shapefile for the entire African continent based on the Excel file; and as files for particular event aggregations, included “Civil war” (events associated with government and rebel battles and all rebel activity), or “Violence against civilians”. Please note that, while xls or csv sheets can be opened by most software data packages, the shapefile will require a GIS package or others than can read spatial data (e.g. R).

Additional information on ACLED methodology and on how to download ACLED data files can be found online in our codebook and user guide.

 


What is the difference between the monadic and dyadic files?

Most data analysis can be carried out using the standard dyadic file. In this file, both Actor 1 and Actor 2 appear in the same row, with each event constituting a single unit of analysis. This allows users to analyse overall levels of violence, numbers of events, and the geographic patterns of conflict overall, without specifying an individual actor.

However, in other cases, a monadic file is more useful. This is a file in which Actor 1 and Actor 2 appear in a single column, with each actor’s activity constituting a single unit of analysis. A monadic file allows users to analyse things like the proportion of events in which a particular actor or actor type is involved, or the geographic patterns of activity of specific actors.

The full, annual dataset is already available as both a monadic and dyadic file on our site. Users can also create their own monadic files of the realtime data, in the event that they wish to include the most recent data in their analysis. For a summary of how to create monadic files of realtime data, please see below.

 


How do I create a monadic file of realtime data, or realtime and Version 6 / 7 data combined?

Creating a monadic file involves duplicating the events so that each actor is represented as participating in a single event (almost doubling the number of events in the dataset); then deleting all the events with no actor in the single column (one-sided events).

 


Why does the realtime file contain events from previous periods?

ACLED periodically carries out reviews on historical data (from 1997-2016) as part of our revision process, adding newly coded events from historical periods. Due to occasional reporting lags, and/or insufficient detail in early event reports for inclusion in the dataset, a small number of events in the 2017 data pre-date this period. These events do not duplicate or reproduce any previously recorded events, but are exclusively new additions to the dataset coded in our current coding period.

If any recorded events fall outside the date range of your interest, the data can be filtered or sorted to exclude these earlier events; alternatively, the realtime data can be merged with the full annual dataset to provide the most comprehensive picture of recorded political violence and protest. All events coded in the current calendar year will be included in a fully revised and updated version of the annual dataset in January 2018.

 


Can I use ACLED data to find information about a specific group?

Analysis of particular groups or perpetrators of violence requires the use our monadic file. This is a file in which Actor 1 and Actor 2 appear in a single column, with each actor’s activity constituting a single unit of analysis. This allows users to analyse features like the proportion of events in which a particular actor or actor type is involved, or the geographic patterns of activity of specific actors. Creating a monadic file involves duplicating the events so that each actor is represented as participating in a single event (almost doubling the number of events in the dataset).

 


Can I use ACLED data to find information about the number of people killed by a specific group?

ACLED does not code fatality figures according to which group suffered casualties because most source reports do not offer this level of detail. Instead, the events include, when available, the total number of deaths arising from a conflict event. The estimated number of fatalities associated with a single event is reported in the fatalities column.

For this reason, the data cannot generally be used to estimate the number of deaths caused by one actor or another in a conflict, as a single event may contain information on casualties suffered by both parties in a battle, for example. The only exception to this is in incidents of violence against civilians: because ACLED only codes events of violence against civilians where the targets were unarmed, non-combatants, the number of fatalities reported for each event of violence against civilians is taken to be the reported number of civilians killed.

All fatalities recorded are ‘reported fatalities.’ ACLED does not independently verify details of fatalities, and includes this information as an estimate only, reflecting the content of media reports. We further specify in our user guides and other resources that fatality data are particularly prone to manipulation by armed groups, and occasionally the media, for various reasons, and urge users to take this into account in their analysis of fatalities. As such, ACLED codes the death toll as it is reported; where a range of fatalities is reported, we code the lowest of that range, and seek to note in the ‘Notes’ section when there has been a dispute.


Can I use ACLED data to find information about the number of people injured by a specific group?

The fatalities column in our dataset refers to the estimated number of reported fatalities associated with a single event. ACLED does not collect casualty data, which may include injuries as well, but reported fatality data only.