Social media and digital technologies are changing the way information about political violence is collected, disseminated, analysed and understood. Effective early warning and crisis response increasingly depends on the availability of timely, reliable reports of violence, and a growing body of research on violence relies on the availability of reliable violent event data to understand patterns, dynamics and trajectories of violence. While biases in traditional media – newspapers and print media – have been analysed and documented in the literature, there is relatively little information about biases in relation to new and emerging sources of data. This paper seeks to determine the comparative opportunities and limitations of ‘new’ and ‘old’ data sources for early warning, crisis response, and violence research. We compare the information set produced through social media violence reporting with conventional violence reporting around the August and October 2017 Kenyan elections.
Specifically, we leverage data from a sample of social media reports of violence through public posts to Twitter. These reports are compared with events coded from media and published sources coded by the Armed Conflict Location & Event Data Project (ACLED) along three dimensions: (1) the geography of violence reporting; (2) the temporality of reporting; and (3) the targeting of reporting.