Making uncertainty explicit : separating reports and events in the coding of violence and contention

Abstract

When coding events from media sources – as the majority of data projects do – different reports may oftentimes contain contradictory information. What do coders make of this? It is up to them to aggregate different reports into one coded event, and to supplement missing information based on other sources or their own background information. If not addressed properly, this may lead to a lack of replicability and to low reliability of the final data product. In this short article, we present an approach for separating (i) event reports and the information contained in them, and (ii) events, which are based on aggregate information from the reports and constitute the final data product. Our procedure preserves uncertainty arising from multiple reports and gives the user control over how missing and conflicting information should be dealt with. We illustrate our procedure with data from a current coding project, the Mass Mobilization in Autocracies Database (MMAD).