Tracking Civic Space in Developing Countries with a High-Quality Corpus of Domestic Media and Transformer Models

Civic space - the fundamental freedoms necessary for citizens to influence politics -is under constant contestation. Despite the importance of day-to-day contestation overthese rights, there is very little data allowing us to study the events and processes thatconstitute this struggle. We introduce new data that captures civic space activity across65 developing countries from 2012 to 2024. Using an original corpus of over 120 millionarticles from nearly 350 high-quality domestic media outlets and 30 international andregional outlets, we use human-supervised web scraping and open-source computationaltools to track monthly variation in media attention across 20 civic space events. Ourapproach yields three achievements: first, our corpus provides unprecedented coverageof reporting by developing country media outlets, addressing biases in other mediaevent data; second, the resulting monthly event data set covers a wide range of newcivic space activities; and third, we demonstrate the utility of this data for identifyingand forecasting major political events and discuss applications for research on regimedynamics during a time of democratic backsliding.