Data Description and Methodology
In EDACS, the basic unit of analysis is an event, defined as a violent incident with at least one fatality resulting from the direct use of armed force. EDACS currently contains detailed information on violent events in six countries of Sub-Saharan Africa (Somalia, Democratic Republic Congo, Republic of the Congo, Burundi, Sierra Leone, and Liberia) between 1990 and 2009.
Events are coded with their location (name of location, longitude and latitude coordinates) and timeframe (start and end date in case of events lasting more than one day). Among others, the type of military action (fighting or diverse forms of one-sided violence) is coded, as well as the involved violent or non-violent actors and details on (civilian and military) fatalities. An overview of the various variables provided in the dataset can be found in our Codebook.
Each news article found by the search engine is read by our coders who extract the relevant information and enter it into a data entry mask. In order to ensure inter-subjectivity and consistency of our data, we developed a set of strict and conservative coding rules that can be found in our Codebook. Additionally, all data is coded and double-checked by two different coders and cross-checked by a supervising coder.
Events are localized with longitude and latitude coordinates (WGS 84) using the toponymic GEOnet Names Server (GNS) provided and maintained by the US National Geospatial-Intelligence Agency (NGA). In case of imprecise (“between town A and town B”) or ambiguous (duplicate location names in the GNS data) indications of event locations in our primary sources, we consult additional map data such as GoogleEarth or Harvard’s AfricaMap and/or apply a variety of standardized buffer rules that can be found in our Codebook.
Consulting qualitative conflict reports as well as consolidated actor databases (e.g. UCDP, IISS), actors are coded by their most commonly used name. In case an actor cannot be identified, we translate the information provided by the sources, albeit often biased, to generic actor categories such as “rebels”, “bandits”, “clan militia”, or “unspecified actor”.