This article introduces a novel way to predict armed conflict by analyzing newspaper text. Using machine learning, the text is transformed into interpretable topics that capture changing contexts.
📅 Timing Prediction: The focus shifts from predicting conflict only where it has occurred before to using within-country topic variation.
🔍 Method Insight: Topics provide depth through evolving terms and width via summarizing full content including hidden stabilizers or destabilizers.
📊 Findings Summary:
• Topic modeling extracts meaningful themes from vast newspaper archives
• Panel regressions connect these thematic shifts with conflict onset data
• This approach successfully identifies emerging risks in peaceful countries
🌐 Broader Relevance: It offers a fresh perspective on political violence prediction, highlighting the power of textual analysis beyond traditional indicators.
| Reading Between the Lines: Prediction of Political Violence Using Newspaper Text was authored by Hannes Mueller and Christopher Rauh. It was published by Cambridge in APSR in 2017. |
