Fleeing from Hurricane Irma: Empirical Analysis of Evacuation Behavior Using Discrete Choice Theory

Abstract

In 2017, the United States (U.S.) was severely impacted by a number of devastating natural disasters that required mass evacuations. Of these disasters, Hurricane Irma in September 2017 led to one of the largest evacuations in U.S. history, involving over six million people. This research analyzes the behavior of a survey sample of evacuees and non-evacuees (n= 645) from Hurricane Irma (September 2017) by developing advanced modeling techniques in discrete choice theory. Given recent increases in size and scope of natural disasters and evacuations around the world, this work is a timely contribution to the evacuation and emergency management literature. We uncover and solve several key gaps in the literature related to evacuee behavior by 1) modeling the full multi-dimensional and interdependent nature of evacuee choices; 2) developing a portfolio choice model for evacuation behavior; and 3) advancing a latent class choice model using revealedpreference hurricane evacuation behavioral data. We use these models to infer new behavioral insights from our data, which can be leveraged by practitioners to improve preparedness and response strategies for large-scale evacuations.

Words: 2949 (not including abstract)