In data we trust? A comparison of UCDP GED and ACLED conflict events datasets

Abstract

In recent years, several large-scale data-collection projects have produced georeferenced, disaggregated events-level conflict data which can aid researchers in studying the microlevel dynamics of civil war. This article describes the differences between the two leading conflict events datasets, the Uppsala Conflict Data Program Georeferenced Events Dataset (UCDP GED) and the Armed Conflict Location Events Dataset (ACLED), including their relative strengths and weaknesses. The aim of the article is to provide readers with some guidelines as to when these datasets should be used and when they should be avoided; it finds that those interested in subnational analyses of conflict should be wary of ACLED’s data because of uneven quality-control issues which can result in biased findings if left unchecked by the researcher. The article concludes that those interested in non-violent events such as troop movements have only ACLED to choose from, since UCDP has not coded such data, but again warns researchers to be wary of the quality of the data. Finally, while the creation of these datasets is a positive development, some caveats are raised in relation to both datasets about the reliance on media sources.