ACLED Event Analysis
- Introduced by
- Clionadh Raleigh, Andrew Linke, Håvard Hegre & Joakim Karlsen
- Year
- 2010
- Type
- Disaggregated coding and analysis of political-violence events
- Family
- Conflict event-data analysis
What does it do?
ACLED event analysis is the disaggregated study of political violence and protest using the Armed Conflict Location and Event Data project, introduced by Raleigh, Linke, Hegre, and Karlsen (2010).
ACLED codes individual events — battles, violence against civilians, riots, protests, explosions and remote violence, and strategic developments — with their date, location, actors, and any fatalities, updated on a near-weekly basis. Its fine granularity and timeliness make it a workhorse for mapping, monitoring, and modeling where, when, and by whom political violence occurs.
The intuition behind the method
Country-year counts of war hide almost everything interesting about conflict: a single 'civil war' is really thousands of distinct events scattered unevenly across towns and weeks, mixing pitched battles, massacres of civilians, riots, and peaceful protests turned violent. ACLED zooms all the way in, recording each event as a dated, mapped incident with named perpetrators and victims and a typology that keeps a protest distinct from a battle. Because it is updated almost in real time, it can power a live dashboard of unrest as well as a careful statistical study. The trade-off for this breadth and speed is that it casts a wide net — including non-fatal events — so analysts must understand its coding choices to use it well.
History
ACLED was launched by Clionadh Raleigh and collaborators and introduced in a 2010 Journal of Peace Research article, initially covering African conflicts before expanding to near-global coverage. Designed for disaggregated, real-time data on political violence and protest, it grew into one of the most widely used conflict datasets in both academic research and the policy and humanitarian communities, complementing the more conservative, fatality-gated UCDP data.
When to use it
Use ACLED when you need broad, timely, finely disaggregated data on political violence and protest — including non-fatal events and contentious politics — for monitoring, mapping, or high-resolution spatial-temporal modeling, particularly in recent periods and across many regions. It is less suited when you require a long historical series with strict comparability (UCDP or COW are better), when only fatal organized violence matters, or when the wide inclusion net and source-driven coverage changes would confound the analysis without careful handling.
Limitations and considerations
- Wide inclusion rules and source dependence create reporting and media biases and coverage that changes over time.
- Fatality figures are estimates and less conservatively coded than fatality-gated datasets.
- Expanding coverage and sourcing over time can create artificial trends if not handled carefully.
- Event counts conflate very different acts unless event types are kept distinct.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Fine-grained, georeferenced, dated events with a rich actor and event typology.
- Near-real-time updating supports monitoring, early warning, and rapid analysis.
- Broad scope including non-fatal violence, riots, and protests beyond armed conflict.
- Wide and expanding geographic coverage with public dashboards and downloadable data.
Applications
- Real-time monitoring and early warning of political violence and unrest.
- High-resolution spatial-temporal modeling of where and when violence occurs.
- Studying protest dynamics and the escalation from contention to violence.
- Humanitarian and policy situational awareness via interactive dashboards.
- Providing disaggregated outcome data for conflict-forecasting systems.
How it works
How it works
ACLED organizes political violence and disorder into a fixed set of event types — battles, violence against civilians, riots, protests, explosions/remote violence, and strategic developments — each with defined sub-event types. This typology lets analysts separate armed combat from civilian targeting from contentious but non-violent mobilization.
Formula
\text{type} \in \{\text{battles, VAC, riots, protests, explosions, strategic developments}\}Trained coders extract events from a wide range of sources — local and international media, NGO and partner reports — recording the date, the two actors involved (with a detailed actor and actor-type coding), the precise location with latitude and longitude and a geographic precision code, and a fatality estimate. Non-fatal events are included, unlike fatality-gated datasets.
Formula
e = (\text{date}, \text{actor}_1, \text{actor}_2, \text{type}, \text{lat/lon}, \text{fatalities})Because every event is georeferenced and dated, events are aggregated to whatever spatial-temporal unit suits the question — administrative units or grid cells crossed with weeks or months — producing panels of event counts, fatalities, or type-specific activity over space and time.
Formula
\text{aggregate to admin unit / grid} \times \text{week or month}The aggregated panels feed spatial and time-series models of violence, early-warning dashboards, and descriptive monitoring. Analysts model counts or onsets as functions of local covariates, exploit the near-real-time updates for forecasting, and visualize hotspots and trends directly from the geocoded events.
Formula
Y_{gt} = f(\text{covariates}_{gt}),\ \text{spatial-temporal models / dashboards}Common pitfalls
- Interpreting rising event counts as rising violence when the increase reflects expanded coverage or sourcing.
- Pooling battles, protests, and strategic developments into one 'conflict' measure, losing crucial distinctions.
- Comparing ACLED totals directly to UCDP or GTD without reconciling definitions and inclusion criteria.
- Treating geographic precision codes as uniform when many events are located only to an approximate area.
Frequently asked questions
How does ACLED differ from UCDP?
ACLED codes a broader set of events — including non-fatal violence, riots, and protests — and updates near-weekly with extensive actor detail, prioritizing breadth and timeliness. UCDP requires at least one fatality per event, applies conservative sourcing, and emphasizes long-run comparability. ACLED captures more low-level and contentious activity; UCDP is more conservative and historically deep. Many analysts use them for complementary purposes.
Why does ACLED include protests and non-violent events?
Because political disorder is a continuum: protests can escalate into riots and violence, and tracking the full spectrum of contentious politics lets analysts study escalation and de-escalation rather than only the violent endpoint. Researchers can always filter to violent or fatal event types, but the broader coding supports questions about mobilization and unrest that fatality-gated data cannot answer.
What is the biggest analytical risk when using ACLED over time?
Coverage and sourcing have expanded as the project grew and added regions and partners, so apparent increases in events can reflect better data collection rather than more violence. Analysts must account for this — for example by restricting to consistently covered places and periods, modeling coverage explicitly, or focusing on within-unit changes — to avoid mistaking measurement expansion for real trends.
Related methods
Sources
Raleigh, C., Linke, A., Hegre, H., & Karlsen, J. (2010). Introducing ACLED: An armed conflict location and event dataset. Journal of Peace Research, 47(5), 651–660.
Open DOI referenceCopied.