Bridging Twitter and Survey Data for Evacuation Assessment of Hurricane Matthew and Hurricane Irma

Abstract

Evacuations are the most effective protective strategy adopted to minimize the deadly threat of an incoming hurricane. The study of evacuations has a long history in the United States, and the scientific community has acquired an in-depth understanding of the associated processes. However, there are limitations in the traditional methods of studying evacuation behavior, such as survey questionnaires and traffic counts, which sometimes fail to efficiently produce the results that emergency managers need. The development of innovative methods for the study of evacuation decision-making, such as the use of so-called big data analytics, offers an opportunity to complement traditional approaches in evacuation behavior research as well as advancing understudied aspects of evacuation behavior. Applied to Hurricane Matthew in South Carolina and Hurricane Irma in Florida, the findings presented in this article reveal that geotagged tweets provide a representative sample of individuals in the age range of 18–54 years, complementing elderly-biased questionnaire survey samples. The combination of questionnaire and Twitter methods found no gender effect on evacuation decision-making but confirmed a significant positive association with residential status. Additionally, the authors observed differential behavior with respect to race/ethnicity and age predictors in the two areas of study. The validity of Twitter is confirmed in connection with exploring certain aspects of evacuation behavior, illustrating a new and rich data source for researchers in this field.