It will rain: The effect of information on flood preparedness in urban Mozambique

Floods are among the costliest and most recurring natural hazards. Many governments provide disaster preparedness information, but there is limited evidence of its effectiveness. Using a randomized experiment across 300 urban communities in Mozambique, I show that providing information on flood risk increased the implementation of suggested mitigation strategies, measured objectively (using a machine learning-based algorithm to detect solid waste, a major flood hazard, in over 25,000 photos) and through self-reported behaviors. Sending flood preparedness text messages proved to be at least as effective as more resource-intensive interventions that involved household visits and video screenings. Risk perception increased, particularly in lower-risk areas, suggesting that information and past experiences contribute to community preparedness by shaping perceived flood risk. These findings highlight the importance of communication in mobilizing communities for disaster risk management, even in contexts where flood risks are presumably already well-known, and underscore the potential of low-cost, scalable communication tools. • Flood preparedness information raised mitigation behavior in urban Mozambique. • Text messages were as effective as costlier video interventions. • Solid waste near drainage canals fell by 6–13 percentage points. • Machine learning from 25k photos provided objective behavior measures. • Low-cost communication tools can enhance urban climate resilience.