No abstract

The United States' National Environmental Policy Act (NEPA) requires federal agencies to use environmental impact assessment tools to aggregate scientific information, garner stakeholder input, and weigh alternatives for infrastructure siting and operations decisions.Thus, NEPA plays a sweeping role in federal infrastructure provisioning. Using a comprehensive dataset of Environmental Impact Statements (EISs) prepared under NEPA from 2013 to 2019, we use natural language processing to identify geographic locations and artificial neural networks to classify discussed environmental impacts. We then describe geographic, temporal, and environmental trends in federal infrastructure provisioning. Finally, we use Bayesian hierarchical regression to assess how socioeconomic and demographic covariates shape county-level EIS counts. Numerous federal agencies are involved in EIS production, but three agencies author the majority of EISs. Total EISs issued annually have declined since the early 2000s. EISs are more frequent in urban coastal counties, but do not show clear distributive trends by income or race.