Quick Guide to ACLED Data
Last updated: March 2023
This Quick Guide provides a brief overview of the most frequently asked questions by users of ACLED data. More detailed information on the topics below can be found in additional methodology documents, all of which are available in the ACLED Resource Library. If you have specific questions that have not been addressed in our guides, please feel free to contact us via email at [email protected].
- What are ACLED data?
- What can ACLED data be used for? What should they not be used for?
- How can I access and use ACLED data?
- How can I find data on a specific theme, such as violence targeting civilians and women, or disorder related to regional trends ?
- How do I cite ACLED data and/or other ACLED publications?
- How are the data structured?
- How are events disaggregated?
- What is ACLED’s geographic and temporal coverage?
- Is historical ACLED data ever revised?
- How is information collected?
- How are the data coded?
- How does ACLED ensure that the data are comparable across regions and contexts?
- How is the quality of ACLED data ensured?
- How can I analyze ACLED data?
- What types of events does ACLED code?
- What is the ‘disorder type’ variable?
- How precise are the coded dates and locations?
- How are injuries and fatalities recorded?
- Is it possible to identify the number of people killed by a specific group or the total number of civilians killed?
- Why doesn’t ACLED have a ‘crowd size’ variable and what does the ‘crowd size’ tag in demonstration events capture?
- Why doesn’t ACLED have a ‘reason for demonstration’ variable?
- Can it be assumed that Actor 1 is the aggressor?
- What are interaction codes?
- Where do I find further documentation on methodology and coding decisions
- How can I make a comment or pose a question to the ACLED team?
Quick User Guide
What are ACLED data?
The ACLED dataset contains disaggregated incident information on political violence, demonstrations, and select related non-violent developments around the world, as outlined in the ACLED Codebook. It is an event based dataset. ACLED data detail the event type, involved actors, location, date, and other characteristics of these incidents. ACLED data are collected in real time and published on a weekly basis according to well developed methodological principles, resulting in current, reliable, and comparable data. Data go through multiple rounds of review prior to publication.
What can ACLED data be used for? What should they not be used for?
ACLED data are widely used for trend monitoring, early warning, and risk analysis that supports and informs operational safety and security decision making. ACLED data are used by individuals, media outlets, scholars, policymakers, and practitioners around the world for a wide range of applications, from risk assessments and early warning initiatives to human rights advocacy and academic research.
There are limitations to the application of ACLED data. ACLED data are granular, comprehensive, and near real-time. But these data are not collected or organized to be a primary source for day-to-day operational safety and security monitoring, for example. Data are available down to the day level (e.g. 6 June 2020), but not to a specific time of day an event took place; data are available to the city/village level (e.g. Atlanta, Georgia, USA), rather than street level; in some unique cases, large cities may be disaggregated into the neighborhood level. ACLED data do not contain personally identifiable information (e.g. names of individuals or mobile device IDs), and cannot be used to track individuals.
How can I access and use ACLED data?
How can I find data on a specific theme, such as violence targeting civilians and women, or disorder related to regional trends ?
The various data columns, such as the Actors, Event Type, and Disorder Type columns, can be used to filter events by specific event types and actors. The Tags column can be used to filter events by specific trends that are currently tracked by ACLED. While most tags are used to track regional- or country-specific trends, some (e.g. the set of tags tracking political violence targeting women), are coded across all regions. For the full list of tags currently coded by ACLED, see this document.
The Notes column contains more detailed descriptions that may allow for heuristic filtering based on keywords — though please keep in mind that the Notes column is qualitative and not standardized, meaning that keyword searches may yield false positives and/or may omit true positives. In addition, ACLED provides a range of curated datasets for different regions and themes, such as violence targeting or involving specific actor types. For the full list of ACLED data columns and their descriptions, see the ACLED Codebook.
How do I cite ACLED data and/or other ACLED publications?
How are the data structured?
ACLED data are available in a spreadsheet and each row represents an individual event. There are multiple columns which contain specific information on the location, date, involved actors, fatalities, and other characteristics of the event. Detailed information about the data columns is available here.
How are events disaggregated?
The ACLED dataset provides information about political violence and demonstration events, disaggregated by date (when the event happened), type of violence (what happened), actors (who is involved), and location (where the event happened). Reports of incidents are broken into individual, discrete events. Each event is one that took place at a specific time, involved distinct types of violence and actors, and/or occurred in a particular location. In practical terms, this means that incidents that differ on time, location, agent and/or event type are all coded as separate events. Further, two incidents that occur on the same day and place, involving the same actors and type of activity, will be aggregated into a single event. For more on this coding process, see the ACLED Codebook.
What is ACLED’s geographic and temporal coverage?
ACLED collects data on political violence and demonstration events in all countries and territories around the world. ACLED maintains real-time coverage for all countries and territories, while historical coverage varies by region. A full list of ACLED’s country and time period coverage is available here.
Is historical ACLED data ever revised?
ACLED is a living dataset and its coverage is updated by adding new events and updating existing events as new information becomes available. There is no limit to how far back events are updated within our temporal coverage per country. For more about ACLED’s coding and review process, see here.
ACLED also engages in backcoding projects to maintain coverage consistency with older data. New sources are regularly covered years back before the data is released in order to prevent a temporal coverage bias.
How is information collected?
ACLED researchers systematically collect and review the latest reports from selected local, national and international sources, including media, vetted social media accounts, government and NGO reports, and partner organizations. ACLED researchers work to triangulate reports when and where possible, but they do not independently verify events or gather first-hand information on the ground. ACLED’s local partners often verify and collect first hand information. ACLED employs a range of sourcing strategies to ensure the data are timely and reliable.
If users come across events that they believe may be missing, they are encouraged to reach out to the ACLED team with such information at [email protected].
How are the data coded?
ACLED researchers scrutinize the information collected during our sourcing process and code the relevant variables according to our global methodology and country-specific methodology rules. Researchers frequently engage with ACLED colleagues and managers on the interpretation of events and the appropriate methodology to ensure that collected data represents the situation on the ground accurately. The classifications of event information are coded to the standards and specificity of ACLED global methodology, ensuring cross-country comparability. In complex contexts, ACLED team members engage with external experts and/or partner organizations to inform decision making around coding decisions.
How does ACLED ensure that the data are comparable across regions and contexts?
ACLED has developed a general methodology with principles that apply across all time periods and regions covered in the dataset. This is documented in the ACLED Codebook and in separate explainers that provide additional information about different facets of the core methodology.
In order to accurately capture events in unique local contexts, ACLED develops methodology and coding decisions tailored to specific countries and conflicts. For example, a protest does not look the same in each region. Furthermore, tailored strategies are needed to capture events and deal with inherent biases in sources in different settings. Region-specific guides explain our approach to these local contexts and what specific methodology and coding decisions apply.
How is the quality of ACLED data ensured?
ACLED data are collected and coded by a team of experienced researchers with specific language skills and country knowledge. Review processes oversee the collected data to ensure intra-coder reliability (i.e. that the researcher has coded everything in their country consistently). The data are further investigated and cross-checked by a regional research manager, who reviews inter-coder reliability across a region (i.e. that all researchers have coded information consistently across the region). Lastly, the data go through a final review from the global methodology team to ensure inter-code reliability, accuracy and methodological consistency (i.e. that all researchers across regions have coded everything consistently around the world).
In addition to weekly coding and review of the real-time conflict data, collection of historical data supplements the dataset. Comprehensive data management and quality assurance efforts performed on a regular basis ensures that the data reflect the latest and best information available. To that end, ACLED has an internal data quality assurance team.
How can I analyze ACLED data?
Because of the many applications of ACLED data, there is no ‘best way’ to analyze or use the data. ACLED’s own analysis publications provide a range of examples. A common and easily accessible way of analyzing ACLED data is using a spreadsheet program like MS Excel and transforming the data using a pivot table. This way, event data can be filtered for an actor or country, collated by year or month, and specified by event type – a common starting point for exploring the data.
Beyond spreadsheet programs, GIS software like QGIS or ArcGIS are often used to map events and their properties using the provided coordinates for geospatial analysis. Tableau is frequently used to create engaging visualizations, including graphs and maps.
While there is no ‘best way’ to analyze ACLED data, there are incorrect ways. For example, assuming the Actor1 is the perpetrator would be a misinterpretation. We stress that it is necessary to read our methodology thoroughly to understand ACLED variables and the inferences that can (and cannot) be drawn from them. The ACLED team is happy to assist with cases that are not evident from our methodology documentation. You can reach out with your question or potential use case to [email protected].
What types of events does ACLED code?
ACLED data collect information on six types of events, both violent and non-violent, that constitute political disorder. These include:
- Battles: Violent interactions between two organized armed groups;
- Explosions/Remote violence: An event involving one side using remote weapons (e.g. artillery). These events can be against other armed actors, or used against civilians;
- Violence against civilians: Violent events where an organized armed group deliberately inflicts violence upon unarmed non-combatants;
- Protests: Public demonstrations in which the participants are not violent;
- Riots: Violent events where demonstrators or mobs engage in destructive acts against property and/or disorganized acts of violence against people;
- Strategic developments: Strategically important instances of non-violent activity by conflict actors and other agents within the context of conflict or broader political disorder. These can include recruitment drives, incidents of looting, and arrests are some examples of what may be included under this event type. Note that strategic developments are coded differently from other event types, and hence users must remember that they should be used differently from other event types in analysis.
Within these broad event categories, ACLED codes 25 sub-event types that classify different actions. Sub-event types allow for deeper analysis of events by “type”. Additional information on ACLED event types can be found in the Event Definitions Primer and the ACLED Codebook.
What is the ‘disorder type’ variable?
The ‘disorder type’ variable denotes the type of disorder a particular event falls into; either ‘political violence,’ ‘demonstrations,’ or ‘strategic developments.’ Events are classified as one or more of these disorder types based on the ‘sub-event type.’ Filtering the data on this variable will allow for analysis on specific types of disorder. For the full list of ACLED ‘sub-event types’ and their corresponding ‘disorder types,’ see the ACLED Codebook.
How precise are the coded dates and locations?
ACLED data include relevant details on the date and location, as well as an estimate on the precision of reported time and locations. These estimates are geo-precision and time-precision codes. For more on the coding of locations, dates, and precisions codes, see the ACLED Codebook pages 29-31.
Locations are coded to named populated places, geostrategic locations, natural locations, or neighborhoods of larger cities. Geo-coordinates with four decimals are provided to assist in identifying and mapping named locations to a central point (i.e. a centroid coordinate) within that location. Geo-coordinates do not reflect a more precise location, like a block or street corner, within the named location.
Dates are coded to the particular day the event took place; ACLED does not include information about the time of day that an event occurred.
How are injuries and fatalities recorded?
ACLED includes an estimate of fatalities for each event, including events where no fatalities are reported. A column called ‘fatalities’ indicates the number of reported fatalities associated with a single event. Fatality information, either on counts or attribution, is often biased and can differ across source information for the same event. For this and other reasons, ACLED defers to the most conservative and reliable fatality estimate available. This number refers to the total number of fatalities that were reported, across all sides of the event. Fatalities are not attributed to specific groups (see more on this point here), though such detail may be indicated in the Notes column if reported in the source information. ACLED does not collect data on injuries; reports of ‘casualties’ only are assumed to be injuries (not fatalities) and hence coded as zero fatalities. For more on fatality coding see this primer.
Is it possible to identify the number of people killed by a specific group or the total number of civilians killed?
ACLED does not systematically code fatality figures according to which group suffered the fatalities. Most source reports do not consistently offer this level of detail, and the reliability of such information is variable. Instead, the event information includes (when available) the total number of deaths arising from each event. The estimated number of fatalities associated with a single event is reported in the Fatalities column.
The number of deaths caused by one actor or another in a conflict can be difficult to estimate, as events contain reported information on fatalities suffered by both parties. The only exception to this rule is when civilians are targeted since it is assumed, by definition, that the civilians were unarmed non-combatants and not responsible for any of the fatalities. For more, see this methodology primer on fatality coding.
Why doesn’t ACLED have a ‘crowd size’ variable and what does the ‘crowd size’ tag in demonstration events capture?
The size of demonstrations is a commonly overlooked aspect in reporting. When actually reported on, the estimates of the turnout to demonstration events often differ widely depending on whose claim it was; demonstration organizers may claim inflated crowd sizes to make their cause seem stronger, whereas those opposed to the demonstration may downplay crowd sizes. Moreover, sources across the globe differ in the consistency and specificity at which they report on demonstration sizes. Given the inconsistency and unreliability of how crowd sizes for ‘Protests’ and ‘Riots’ events are reported on, ACLED does not record this information as a distinct variable. However, ACLED researchers will still note the size of the demonstration, as reported, in the ‘Tags’ column of each event; this is written as “crowd size=x,” with “x” denoting the size of the demonstration as reported by the source. If sources do not report on the crowd size, the reported crowd size is recorded as “crowd size=no report”. All ‘Riots’ and ‘Protests’ events include a reported crowd size in the ‘Tags’ column.
Why doesn’t ACLED have a ‘reason for demonstration’ variable?
Why demonstrators choose to engage in a demonstration can be very varied, even within the same demonstration. Reporting of demonstrations cannot always capture all of these nuances. As such, ACLED does not readily code such a variable, though the data will include any information the original source noted in the qualitative ‘Notes’ section of the event. However, while the reason for a demonstration can be more subjective, who is involved is less so. It is often quite clear from reporting what groups, movements, etc. may be taking part in a demonstration. ACLED does capture this information whenever reported by coding such identifiers as the respective ‘Associated Actor’ within events (for more on coding decisions, see the ACLED Codebook).
Can it be assumed that Actor 1 is the aggressor?
Actor1 and Actor2 refer to the involved parties in an event. The designation of groups as Actor1 or Actor2 is not based on any specific criteria. Coding an actor as Actor1 or Actor2 does not imply that they are the aggressor or initiated the action, nor does it imply that either reacted later, suffered more casualties, or was a victim in an event. The Notes column may contain these details if reported in the source information, though this information may be biased.
What are interaction codes?
All actors in ACLED are categorized by actor type, and each type is associated with an Inter code ranging from 1-8. The combination of the two Inter codes (for Actor1 and Actor2) are joined to generate Interactions that represent engagements between specific types of actors. There are many set interactions that can be used to immediately correspond to a specific type of engagement. For example, an interaction code of 12 indicates that state forces interacted with a rebel group. For more information on actors and interaction codes, see the ACLED Codebook page 19-28.
Where do I find further documentation on methodology and coding decisions?
ACLED has developed a methodology for systematically collecting reports and coding the information into structured and reliable event data. This is documented in the ACLED Codebook, which explains the basic principles that apply across the dataset. Separate, in-depth explainers on ACLED’s core methodology provide further detail on specific elements and coding decisions, such as the coding of fatalities or the coding of actor types, like gendered associated actors.
In order to accurately capture events in unique local contexts, ACLED develops methodology and coding decisions tailored to specific countries and conflicts that serve to further refine the core global methodology. Region-specific guides explain our approach to these local contexts and what specific methodology and coding decisions apply.