This project investigates whether international media coverage of humanitarian crises aligns with their measurable severity. Using a dataset of news articles spanning ten crises from 2009 to 2025, we apply exploratory analysis, feature engineering, and regression modeling to identify predictors of monthly media attention. We find that coverage is driven more by the timing of crises and geopolitical salience than by humanitarian scale.
Read the Full Report (PDF)Gaza received 83 times more articles per person affected than Sudan, despite Sudan having nine times more people in need. Regression modeling on 734 monthly observations shows that crisis duration is the strongest negative predictor of coverage, while funding requirements are the strongest positive predictor. The number of people affected shows little predictive power.
Full code, notebooks, and data pipeline are publicly available.
View on GitHub