When crisis strikes, why do some communities utilize evacuation shelters more than others? This mixed methods study draws on a new dataset of almost-daily tallies of evacuees at 660 local shelters following Japan’s 2018 Eastern Iburi Earthquake in Hokkaido to create a large-N time-series cross sectional (TSCS) dataset of local, short-distance evacuation. We pair time-series cross-sectional data models with qualitative comparative analysis (QCA) of nine affected municipalities to examine why some shelters see higher evacuation rates than others. While past studies have used Facebook user data, post-hoc surveys, or ad-hoc roadside interviews to measure evacuation, this study uses meticulously recorded shelter attendance data to draw inferences about evacuation behavior. Controlling for types of shelters, damage levels, infrastructure quality, social vulnerability, governance capacity, and community resources, we find that in affected communities, stronger bridging social ties, especially when aided by linking ties, motivate greater evacuation to shelters. In unaffected communities, stronger bonding and bridging ties encourage potentially unnecessary evacuation, helping spread rumors during blackouts. These results highlight the necessity of clear, transparent communication with the public, and fostering trust in government during crises.

See full paper here.