Working papers

Households are known to buy flood insurance after floods and let policies lapse as memory fades, but those responses are estimated where the underlying hazard does not change. Wildfire changes it: burned terrain raises downstream flood and debris-flow hazard for several years. Because demand is observed after the fire but before any flood, a response reflects the change in hazard rather than salience alone. Using a staggered difference-in-differences design across western U.S. block groups from 2010 to 2024, we find new policies increase where post-fire flood hazard rises, concentrated outside FEMA-designated Special Flood Hazard Areas and more persistent than landscape recovery would suggest.

Climate change is increasing the frequency of extreme natural disasters, which threaten human life and safety. Evacuation is a key protective action against hurricane danger, but it may be costly for individuals. Despite being an important input to emergency management decisions, there is little evidence on the size of these costs. We combine spatial data on hurricanes, flooding, and evacuation orders with millions of records of individuals' cell phone-derived movement activity to study how communities behave during hurricanes. These data generate rich insights about evacuation behavior, including differences by physical risk, information provision, and demographics. Using a structural travel cost model we estimate that the welfare costs of evacuation can be large when compared to mortality costs, averaging a ratio of one to ten but occasionally exceeding the value of statistical life lost.

This paper investigates the role of experienced risk in adaptation decisions. I exploit spatial variation in flooding to estimate the causal effect of Hurricane Sandy on people's decisions to insure against future flood damages. Since the storm, flood insurance demand in flooded areas has continuously increased relative to nearby areas that were not flooded. The estimated insurance response was driven by the purchase and retention of relatively cheaper policies located in the most flood-damaged areas, implying that cost was a critical factor in people's adaptation decisions. Simulated flooding extents of six other recent events give evidence that Hurricane Sandy's adaptation response was the exception and not the rule.

Globalization of production networks means that disasters have impacts far from where they directly hit. Localized damage to facilities and infrastructure can slow or shut down production, causing input scarcity, price distortions and declines in output and revenue for non-local firms connected through the supply chain. This paper tests the hypothesis that diversified trade networks dampen natural disaster shocks that propagate through supply chains. We construct two measures of input supplier specificity, one on the extensive margin of supplier concentration and another on the intensive margin, and find that a diversified supply chain reduces the shock of a natural disaster on both margins.