Political Bias in Disaster Aid: Evidence from Administrative Data and from Space

• In Event: Informing Disaster Risk Management: Analyses of Forecasting, Short-term Displacement, Public Disaster Assistance, and Long-term Relocation

Abstract

As climate change threatens to increase the severity and scope of natural disasters, governments face the challenge of efficiently and fairly distributing increasingly scarce aid to impacted communities. In the United States, this process relies on joint efforts between local and state officials of affected territories, the White House,and bureaucrats within the Federal Emergency Management Agency (FEMA), a process which has often been pointed to as potentially susceptible to political bias.Past studies have suggested that political variables, such as political party alignment between the president and impacted areas, are correlated with the likelihood of a disaster declaration (Reeves, 2011; Husted and Nickerson, 2014; Schneider and Kunze, 2023). Nevertheless, it remains unclear whether such political biases are driven by elected politicians or by bureaucrats. This study examines political bias in the two steps of the disaster declaration and aid allocation process: (1) the FEMA bureaucrat-driven preliminary damage assessments (PDAs) and (2) the subsequent, White House-driven declaration of disaster and dissemination of public assistance (PA) funding. 

We do so by combining a novel dataset of the FEMA PDAs of disaster-stricken areas with measures of flood damage from hydrological models and from satellite imagery. We fail to find evidence of political bias in the first step (FEMA-led damage assessments and their assistance requests), but find party alignment contributes significantly to the second step (White House-led PA disbursal in response to PDAs), even when controlling for these initial assessments by FEMA and for flooding intensity. To the best of our knowledge, this study is the first to exploit the novel data set of PDAs to assess the role of bureaucrats and politicians in the disaster aid process. Further, contrary to most previous studies, we do not rely on self-reported damage estimates (which are prone to systematic measurement errors (Gallagher,2023)) in estimating political bias.

Author

• Hannah Farkas, Columbia University
  Presenting Author