Single-family residential reconstruction after disasters: Building Permits Survey data and synthetic controls

Jack W. Baker, Nikola Blagojević, and Shujun Zhang

Understanding how quickly and completely communities rebuild after disasters is critical for recovery planning, yet systematic measurement of post-disaster reconstruction remains elusive. The study uses U.S. Census Bureau Building Permits Survey data to measure how disasters affect single-family residential new construction. For each disaster-affected place, we construct a synthetic control using weighted combinations of non-disaster places from the Building Permits Survey in the same Census region. Divergence between actual and synthetic permit activity after the event reveals the disaster's causal effect on new construction. We analyze 18 disasters that occurred between 2004 and 2023, including wildfires, hurricanes, and tornadoes. Wildfires produce large, sustained increases in permitting, with a median peak of more than five times the synthetic control, persisting for several years. Tornadoes produce moderate increases of roughly two times the counterfactual. Hurricanes---even those causing hundreds of millions of dollars in housing damage---produce a more modest average increase in new construction permits. We attribute these differences to the nature of the damage: wildfires and tornadoes tend to destroy structures completely, requiring new construction (and a new building permit), whereas hurricanes more often cause repairable damage, so recovery activity is less apparent in new construction data. These findings help quantify the timing and magnitude of new construction across a range of geographies and disaster types, and clarify the scope within which building permits data can serve as a proxy for post-disaster housing recovery.

Average permitting trends: ratio of actual to synthetic permits, aligned on disaster date.The vertical axis is logarithmic; a value of 1 means actual permits match the synthetic control.Each thin line is one place; the bold line is the median. The bottom-right panel shows placebo(non-disaster) places for comparison