Abstract:
Increasing evidence links exposure to extreme weather events in utero with adverse health outcomes at birth, including lower birth weight. This research, however, often faces data limitations because natural disasters may be localized, often affecting some neighborhoods but not others, whereas outcome data are often available only at higher geographic levels, such as counties. In this article, we introduce a novel strategy for estimating the effects of geographically bounded disasters when localized outcome data are unavailable. We employ this strategy to estimate the effect of exposure to severe tornadoes on infant birth weight in the United States from 1991 to 2017. We merge county-month data on singleton births with block-group-level monthly data on the paths of severe tornadoes and block-group data on the distribution of the population at risk of a birth. We then estimate difference-in-differences models in which the treatment variable is equal to the percentage of the population at risk of a birth affected by the tornado. This strategy results in an estimand that is both more interpretable and more policy-relevant than estimands from traditional models. Our findings demonstrate that exposure to a tornado during pregnancy reduced birth weight for Black mothers.