Is Media Sentiment Associated with Future Conflict Events? 



Abstract

These is increasing interest in the potential of media sentiment to be a leading indicator or even predictor of future conflict events. Literature establishes that sentiment can be central to conflict escalation processes, and that news media may capture and reflect peaceful or conflictual sentiment within a given society. Moreover, analysis through machine learning and natural language processing techniques increasingly allow us to gather and process sentiment data at unprecedented scale, depth, and accuracy. We draw on GDELT’s global sample of more than five billion media articles to test the relationship between media reported sentiment and conflict events, utilizing the PRIO-GRID data structure at daily and monthly intervals. We find that more conflictual sentiment is significantly associated with spatially and temporally proximate future conflict events as measured by the ACLED, SCAD and UCDP-GED datasets. We propose that conflict sentiment can help us analyze conflict escalation processes more precisely by measuring emotional intensity and direction through media sentiment analysis, delivering new value for peace and conflict research.

Keywords: Sentiment, conflict, GDELT, GKG, natural language processing, ACLED, media, war, leading indicators, conflict prediction, motivations for violence, machine learning, PRIO-GRID, UCDP

JEL Classification: F51, F23, F52, F61, F63, C55