Job Market Paper: Should We Flee? The Mortality and Economic Impacts of Hurricane Evacuation Orders • [Link to Draft: Updated Oct 9, 2026]
I estimate the mortality and county-level regional economic impacts of hurricane evacuation orders using a novel instrumental variables framework. Relative to similarly affected but non-evacuated areas, evacuated counties experience fewer deaths but lower personal income, employment, and business establishments into the second year following the storm, likely stemming from delayed resettlement in wealthier neighborhoods. I then develop a threshold decision rule that balances the mortality benefits of evacuation against these persistent economic costs and apply this framework to orders issued from 2014–2022. Incorporating persistent economic costs alters policy evaluation as the largest orders generate net benefits that dwarf their costs, but the typical order does not clear the cost–benefit threshold where only about a quarter of recent evacuation orders remain justified under this framework, compared to 64 percent when evaluating mortality benefits alone.
Finalist, Best Student Paper, 2026 Urban Economics Association North American Meeting
Best Student Research, Center for Policy Research End-of-Year Reception (2025)
CPR Working Paper Version: [here]
Reared in Disaster: The Long-Run Harm from Childhood Storm Exposure • [Link To Draft: Updated Jun 14, 2026]
I estimate how exposure to natural disasters in childhood shapes earnings and schooling later in life, using newly assembled daily disaster records dating back to 1951 linked to residential histories in the NLSY79. I compare adult outcomes for children who face similar expected disaster risk but experience different amounts of exposure, finding that severe and repeat disasters reduce earnings. I show that storms between ages 5-10 most reduce educational attainment while those between ages 10-15 cause large wage penalties in adulthood. Finally, Black individuals suffer larger earnings losses than white peers, pointing to inequality from disaster exposure, and suggesting that disaster exposure may have contributed substantially to the racial wealth gap.
Revise & Resubmit, Journal of Urban Economics
CPR Working Paper Version: [here]
Feed Lots and Learning Outcomes (With Maria Zhu)
We provide the first estimates of how concentrated animal feeding operations (CAFOs) affect nearby students. Between 1993 and 1997, more than 2,000 lagoon-based swine operations opened in North Carolina before a statewide moratorium halted new construction. Linking satellite-detected opening dates to 1.8 million student-year test records in grades 3–8, we estimate a staggered difference-in-differences design with student fixed effects. A new operation within two miles of a school lowers math achievement by 0.04 standard deviations, roughly a tenth of a school year's learning, with no detectable effect on reading. Effects diminish with distance, and beyond four miles we estimate precise zeros. Students within two miles of new operations are disproportionately Black, 30 percent compared with 21 percent statewide.
Runner Up, Best Student Research, Center for Policy Research End-of-Year Reception (2026) [Link to Poster]
Nowhere to Return: Housing Loss, Displacement, and Entry into Homelessness After Hurricanes
I estimate the effect of hurricanes on homelessness by linking storm wind fields to the two federal systems that count the homeless. School districts must identify students without fixed housing, including families doubled up or staying in motels; HUD's point-in-time count enumerates observed homelessness on a single January night. Using difference-in-differences designs built for recurring shocks, I find that a hurricane-force landfall raises homeless-student counts by roughly 20 percent in the landfall school year, while point-in-time counts are essentially unchanged. The displacement hurricanes cause takes precisely the forms the point-in-time count cannot see — doubled-up and motel stays — suggesting that the measure guiding federal homelessness policy misses most storm-driven housing loss.
Does Disaster Aid Create Lock-In? Evidence on Recovery and Persistence
Federal recovery aid is meant to help communities rebuild, but rebuilding in place can also anchor households and businesses to locations whose hazard risk has just been revealed. This project asks whether Community Development Block Grant–Disaster Recovery (CDBG-DR) funding produces such lock-in. Identification comes from HUD's "most impacted and distressed" designations: HUD assigns the designation using damage-based thresholds and requires that the large majority of each grant be spent in designated areas, so places on either side of the cutoff experience similar disasters but sharply different aid intensity. Comparing counties and ZIP codes just above and below the threshold, I estimate the effect of aid on out-migration, rebuilding and new construction in high-risk areas, home prices, business survival, and flood insurance take-up, and on exposure and damages in subsequent storms. The results speak to a central tension in disaster policy, whether aid that speeds recovery also raises the cost of the next disaster.
In addition to the projects listed above, I have early-stage projects on:
Does Insurance Lead to Regional Growth? Hurricane Deductibles and Recovery after Disasters [Link to Proposal]
PFAS exposure and student outcomes with evidence from Cape Fear
Immigration enforcement around natural disasters with implications for the cost of recovery