Public Investment in Hazard Mitigation: Effectiveness and the Role of Community Diversity

Abstract

I estimate the loss-reducing effect of local public investments against natural hazards with new measures of damages, weather risk, and spending for a panel of 904 US coastal counties in 2000-2020. I distinguish federally- and county-funded projects and rely on a quasi-experimental strategy, matching counties by economic development, population, and weather risk. Risk predictions come from the Random Forest learning algorithm, using granular data on resident vulnerability and severe weather frequency. Public spending on adaptation is effective -- the average high-spending county avoids a significant portion of losses -- and efficient -- $1 prevents up to $3 in losses over 20 years. The evidence suggests that federal spending is focused on high-risk areas, while local spending is effectively implemented in medium-risk counties. Finally, I show that fractionalization among residents about the priority of climate-change policy can be a limiting factor in adaptation spending. Total spending is significantly lower in areas with high diversity in policy preferences, and more so when opinions are equally split.

Keywords: Natural Disasters, Machine Learning, Resilience, Public Expenditures, County Budget

JEL Classification: H4, Q5, R5