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

This paper analyzes the observed decision-making behavior of a sample of individuals impacted by  Hurricane  Irmain2017(n  =  645)  by applying advanced  methods  based  in  discrete  choice theory.  Our  first contribution  is identifyingpopulation  segments  with  distinct  behavior  by constructinga latent class choice model for the choice whether to evacuate or not. We find two latent  segments  distinguished  by  demographics  and  risk  perception  that  tend  to  be  either evacuation-keen or evacuation-reluctant and respond differently to mandatory evacuation orders.Evacuees subsequently face a multi-dimensional choice composed of concurrent decisions of their departure  day,  departure  time  of  day,  destination,  shelter  type,  transportation  mode,  and  route. While these concurrent  decisions  are often analyzed  in  isolation,  our secondcontribution  is  the development   of   a   portfolio   choice   model(PCM),   which   captures   decision-dimensional dependency(if present)without requiring choices to be correlated or sequential. APCMreframes the  choice  set  as  a  bundle  of  concurrent  decision  dimensions, allowingfor  flexible  and  simple parameter estimation. Estimated models reveal subtle yet intuitive relations, creatingnew policy implications  based on dimensional  variables,  secondary  interactions,  demographics,and  risk-perception variables. For example, we find joint preferences for early-nighttime evacuations (i.e., evacuations more than three days beforelandfall and between 6:00 pm and5:59 am) and early-highwayevacuations  (i.e.,  evacuations  more  than  threedays beforelandfall  and  on  a  route composed  of  at  least  50%  highways).  These  results indicatethat  transportation  agencies shouldhave  the  capabilities  and  resources  to  manage  significant  nighttime  traffic  along  highways  well before hurricane landfall.