The higher the better? The limits of analytical resolution in conflict event datasets

The majority of conflict event datasets rely on media reports as their sole source of information. Because of the various difficulties associated with media reports, it is useful to compare conflict events based on them with those obtained from other observers. A paper published in 2010 by O’Loughlin and colleagues makes a first attempt to do this by using (1) a media-based event dataset and (2) military records on Afghanistan. While the authors conclude that the level of agreement between the two datasets is high, my results show that this goes away once we aggregate to finer analytical resolutions – those that are typically used in micro-level conflict analyses. Thus, rather than giving us the ‘all-clear’ for the accuracy and quality of media-based conflict data, my results once again point to the importance of robustness tests in quantitative conflict research, but also to the need to study the discrepancies in different reporting mechanisms to find out what they can and what they cannot tell us.