Since its original introduction as a small-scale academic project in 2010 covering political violence in Africa (Raleigh et. al, 2010), the Armed Conflict Location & Event Data Project (ACLED) has evolved into an independent NGO with a global team of researchers collecting real-time data on political violence and demonstrations across the world. By mid-2020, nearly one million events were collected in the ACLED dataset for public use, complemented by a range of analytical reports and methodological advancements. Over the past 10 years, a number of significant changes have enhanced the flexibility, timeliness, and reliability of ACLED data. The changes adopted have advanced the project’s scope, deepened its engagement with researchers, and expanded its sources and quality of information. This special data feature reviews these changes and explains why ACLED has evolved in a distinct way compared to other conflict data projects. As a result of this evolution, the ACLED dataset has come closer to comprehensively and accurately representing the ground reality of global disorder than ever before, making it a key source for academics as well as policymakers, practitioners, and journalists around the world.
New Developments in ACLED Coverage
In 2020, ACLED is nearly global in its coverage. ACLED’s geographic expansions have required extensive reflection on how to compare conflicts occurring in different spaces and times; what is ‘political’ violence; and how sourcing and reporting can emphasize or obscure local patterns. These questions underscore an underappreciated fact within conflict studies: that all countries experience some forms of disorder. The modality of that disorder is strongly shaped by the specific vulnerabilities, political fault lines, and priorities of each society. In this article, we review the main changes in ACLED over the past 10 years, concentrating on the breadth and depth of content, sourcing, and methodological advancement.
First and foremost, ACLED expanded where and how it collects data on disorder, which combines political violence, conflict, and demonstrations. ACLED covers disorder wherever and whenever it occurs, rather than relying on the presence and intensity of specific forms of conflict and unrest to spur coverage. Conflict scholars require data where inclusion criteria have not biased the potential research outcomes. Countries that are not typically included in conflict datasets – such as the United States, Ireland, or Botswana – have protests, riots, or acts of repression by state security services. Our most advanced theories of conflict suggest that waning institutional trust and strength, income inequality, and demographics are sources of disorder and causes of political violence (Goldstone, 2001). Given that these features are now commonplace in higher-income states, tracing what may be early signs of instability is in the interest of the research community.
Further, disorder does not present as the same in every space, and by taking an expansive view of its modalities, researchers can use ACLED to accurately account for its occurrence and patterns. Spaces where political violence is high but dismissed as ‘criminal’ or difficult to categorize – such as in Mexico and Brazil – are given equal coverage to more ‘traditional’ conflict spaces, like Afghanistan or the Democratic Republic of Congo. Given existing and developing patterns of instability across the world, it is critical for researchers to consider the modality of unrest in high-income states; the differences in communal violence across identity groups; and the presence and range of armed, organized groups in ‘democratic’ states.
Second, ACLED is a ‘living dataset.’ Data are published on a weekly release schedule; this timing is designed to harness the most accurate and available information, and allows for near-real-time monitoring by users. It is also the basis for a robust and reliable early warning system. Further, ACLED is researcher-led, and prior to publication, data go through multiple rounds of review to account for inter-coder, intra-coder, and inter-code reliability that together allow for cross-context comparability. This regular oversight ensures the accuracy of events, and the avoidance of false positives and duplicates. Oversight continues after data are published: as more information comes to light, corrections and updates are uploaded to the dataset.1 Each week, approximately 10% of submitted events are updates to existing events. Further, targeted supplementation of key periods or through access to new sources of information means that coverage is always being strengthened.2 Third, and finally, ACLED enhanced event data methodologies for manual, researcher-led data collection. ACLED’s innovations here are to first emphasize that a standard ‘one-size-fits-all’ approach to data collection and event coding is not a suitable or adequate response to the variability of modern disorder. Each state and its disorder risks are regularly reviewed by country experts and ACLED team members to ensure that data accurately reflect the most reliable information available. ACLED practices also prioritize data collection by rigorous, human researchers, as opposed to relying mainly or dominantly on automation. This ensures reliable and consistent data collection, and allows ACLED to harness information from a range of languages and sources.3 Further, open, continual, and transparent reviews and discussions of collection processes are features of all ACLED work, and this benefits both data developers and user communities. Users whose research is affected by coding decisions should be informed about how these choices are made, and how they will impact analysis. To fulfill these principles, in addition to codebooks, training, and explanatory materials, ACLED provides open access to methodology documents on a range of significant and specific issues. Examples include obstacles around coverage in Yemen, coding rules for various factions of Boko Haram, and how political violence targeting women data are collected, amongst many others. This methodological transparency ensures that coding remains standardized across countries, conflicts, and researchers. In addition to this documentation, the ACLED team remains accessible, with questions from users regularly received and addressed. These reviews and developments are necessary as even similar conflicts in different regions can be fundamentally distinct, and ongoing methodological decisions are made, implemented, and publicized to inform users. Offering a depository of information from which to make informed choices about event data forms and comparisons is a necessary feature of modern projects.
ACLED captures disorder, defined as a catchment of both political violence and demonstration events. ACLED’s inclusion criteria for a political violence event is that an episode must involve at least one violent group with a political objective. ‘Political violence’ is defined as the use of force by a group with political motivations or objectives, which can include replacing an agent or system of government; the protection, harassment, or repression of identity groups, the general population, and political groups/organizations; the destruction of safe, secure, public spaces; elite competition; contests to wield authority over an area or community; mob violence; etc. ACLED casts a wide net to capture demonstrations as well, allowing any reported violent riot or peaceful protest of more than three people to be included, regardless of the reported cause of the action. Demonstrations are not considered to be ‘political violence,’ yet are a political act. These definitions are important to clarify as an act of political violence or a demonstration is the sole basis for inclusion within the ACLED dataset. Specific non-violent events are also included as ‘strategic developments,’ which are useful to capture contextually important events and developments which may contribute to a state’s political history and/or may trigger future political violence and/or demonstrations.
ACLED’s broad catchment reflects the following tenets: modern disorder forms are best categorized by where they fall on a spectrum of armed, active organization. At one end, a state or international military body with an armed, hierarchical order can engage with all other conflict agents, be present across a territory, and is expected to be perpetually active. In turn, their event forms, intensity, and frequency is extensive. On the other end, there are unarmed protesters engaging in intermittent, highly specific and local demonstrations often with little organization, funding, and capabilities. Between these two extremes are rebel organizations which seek to address national-level power disparities; militias who are ‘hired’ and whose objectives are closely tied to a political patron; community armed groups who provide specific security services on a club good basis; cartels who require territorial and population control to engage in an extensive economic racket; violent mobs who respond to community threats with targeted violence before disbanding, largely emerging in areas with poor police presence and trust; and rioters who emerge to quickly create disorder before reconstituting or fading quickly. The underlying theme to these groups and forms of disorder is that they are all organized to varying degrees in a defined time and space to pursue political objectives, and many arm themselves to do so.