Elsevier

Journal of Choice Modelling

Volume 25, December 2017, Pages 50-60
Journal of Choice Modelling

Modeling joint evacuation decisions in social networks: The case of Hurricane Sandy

https://doi.org/10.1016/j.jocm.2017.02.002Get rights and content

Abstract

Coastal areas of the United States are vulnerable to substantial loss of lives and property damage from repeatedly occurring hurricanes and evacuation is the usual recourse to prevent loss of life when high storm surge threatens. The fundamental question in evacuation modeling is to explore the complex evacuation decision-making process leading to an individual’s decision to evacuate or not during a hurricane threat. Recent studies suggest that the social network characteristics of individuals could potentially determine overall evacuation patterns. This study explores the joint evacuation decisions of individuals in personal networks by using ego-centric social network data obtained from Hurricane Sandy and by considering the nested structure of the ego-centric network data, i.e. close contacts (alters) as nested within an individual (ego). In this regard, the study develops a multinomial multilevel model of joint evacuation decisions at the dyadic (ego-alter tie) level utilizing a Hierarchical Generalized Linear Modeling (HGLM) approach. Model estimation results suggest factors that define a social tie (contact frequency, discussion topic and geographic proximity) significantly influence the evacuation decisions between individuals and their social partners. In addition, individuals’ (both ego and alter) own socio-demographics such as age, marital status, previous evacuation experience, evacuation order, household’s type, size, location and proximity to a water body also affect the decision to evacuate. These findings are useful to help emergency managers implement efficient evacuation strategies and to facilitate planning by policymakers by determining fractions of people evacuating or not for a major hurricane within the context of their social networks.

Section snippets

Introduction and motivation

Hurricane Sandy (October 2012) caused about 254 deaths in the US, Caribbean, and Bahamas with an estimated economic loss of $65 billion (USD) (Ehrhart et al., 2012). Seventy-two fatalities, including 41 drownings from storm surge, resulted from Sandy’s wind and flood impact (Gladwin et al., 2013). Timely evacuation could have averted such storm surge deaths that were higher than all but two prior hurricanes (Katrina and Camille) in the past 60 years (National Hurricane Center, 2015). Most of

Background and related work

Evacuation is a typical form of travel during extreme events and it is the usual recourse to prevent loss of life if high storm surge occurs. However, evacuees often show synchronized behavior in terms of when they depart and what route they take. (Wolshon, 2002). Several studies show important factors affecting evacuation decisions such as hurricane trajectory and warning system, characteristics of evacuees and their households, etc. to explain the overall evacuation process (Baker, 1991,

Data collection

In this study, data was collected by surveying residents from high storm surge risk areas of New York and New Jersey all of whom participated in the survey voluntarily (see Fig. 1). The survey was funded by the National Science Foundation. The sample frame for the survey was based on US census block group areas in coastal New Jersey and New York likely to be at risk of life-threatening storm surge in a major hurricane; the block groups within five km of the coastline or inland tidal water body

Social networks, social influence, and evacuation decisions

Social networks include individuals as nodes and their relationships as links or edges. Such relationships help to construct a network of individuals (Jackson, 2010) and guide how the process of social influence takes place between peers. Studies have suggested that sources of influence may include personal influence between the direct contacts, influence between social groups or social circles, and influence from information sources such as media (Kadushin, 2012). As such, there exists a

Modeling framework

In this study, a Hierarchical Generalized Linear Modeling (HGLM) approach for a multinomial model is used to assess the joint evacuation decision outcome during a major hurricane at the dyadic (ego-alter) level. The multinomial multilevel modeling approach is appropriate for clustering data with a multinomial outcome considering the following two specific reasons: (1) multinomial analysis is more efficient than a series of binary logit analyses since the former approach involves a large number

Model estimation results

Following the modeling framework discussed above, a multi-level multinomial logit model is estimated in this study. As presented in Table 2, all of the variables included in the mixed logit model are statistically significant with plausible signs at the usual 5% or 10% levels of significance. Prior to estimating the full, two-level multinomial HGLM, to examine how much variation in evacuation decision exists between different egos, the empty model, which has no explanatory variable, was

Conclusions and key findings

In this study, the role of social ties on evacuation behavior is explored by using ego-centric social network data obtained from Hurricane Sandy and by considering the hierarchical structure of the data i.e. friends or close contacts (alters) being nested within an individual (ego). With the help of Hierarchical Generalized Linear Modeling (HGLM) technique, this study develops a multinomial multilevel model of joint evacuation-decision outcome based at the dyadic (ego-alter tie) level.

Acknowledgements

The authors are grateful to National Science Foundation for the Grant CMMI-1131503 to support the research presented in this paper. Several questions used for the survey questionnaire were derived from earlier research on the Hurricane Sandy evacuation done by Hugh Gladwin and Betty Morrow supported by National Science Foundation grant CMMI-1322088. The survey was conducted by Hugh Gladwin of Florida International University. The writers also acknowledge Prof. Sharon Christ of Purdue University

References (78)

  • A.M. Sadri et al.

    A random parameter ordered probit model to understand the mobilization time during hurricane evacuation

    Transp. Res. Part C: Emerg. Technol.

    (2013)
  • A.M. Sadri et al.

    Analysis of hurricane evacuee mode choice behavior

    Transp. Res. Part C: Emerg. Technol.

    (2014)
  • H. Sherwin et al.

    An exploration of the importance of social influence in the decision to start bicycling in England

    Transp. Res. Part A: Policy Pract.

    (2014)
  • E. Urbina et al.

    National review of hurricane evacuation plans and policies: a comparison and contrast of state practices

    Transp. Res. Part A: Policy Pract.

    (2003)
  • M.A. Van Duijn et al.

    Multilevel analysis of personal networks as dependent variables

    Soc. Netw.

    (1999)
  • J.C. Whitehead et al.

    Heading for higher ground: factors affecting real and hypothetical hurricane evacuation behavior

    Glob. Environ. Change Part B: Environ. Hazards

    (2000)
  • D.P. Aldrich

    Building Resilience: Social Capital in Post-disaster Recovery

    (2012)
  • E.J. Baker

    Hurricane evacuation behavior

    Int. J. Mass Emerg. Disasters

    (1991)
  • S.P. Borgatti et al.

    On network theory

    Organ. Sci.

    (2011)
  • R.S. Burt

    General social survey network items

    Connections

    (1985)
  • Carnegie, J., Deka, D., 2010. Using hypothetical disaster scenarios to predict evacuation behavioral response,...
  • Carrasco, J.A., Miller, E.J., 2007. The Social Dimension in Action: A Multilevel, Personal Network Model of Social...
  • J.A. Carrasco et al.

    Affective personal networks versus daily contacts: analyzing different name generators in an social activity-travel behaviour context

    Transp. Surv. Method.: Best. Pract. Decis. Mak.

    (2013)
  • Cheng, G., Wilmot, C.G., Baker, E.J., 2008. A destination choice model for hurricane evacuation, In: Proceedings of the...
  • R.A. Clifford

    The Rio Grande flood: a comparative study of border communities in disaster

    Natl. Acad. Sci. Natl. Res. Counc.

    (1956)
  • J.S. Coleman

    Social capital in the creation of human capital

    Am. J. Sociol.

    (1988)
  • N. Dash et al.

    Evacuation decision making and behavioral responses: Individual and household

    Nat. Hazards Rev.

    (2007)
  • Deka, D., Carnegie, J., 2010. Analyzing evacuation behavior of transportation-disadvantaged populations in northern New...
  • V. Dixit et al.

    Understanding the impact of a recent hurricane on mobilization time during a subsequent hurricane

    Transp. Res. Rec.: J. Transp. Res. Board

    (2008)
  • V.V. Dixit et al.

    Strategies to improve dissipation into destination networks using macroscopic network flow models

    Fundam. Diagr. Traffic Flow. Theory

    (2011)
  • T.E. Drabek et al.

    Families in disaster: reactions and relatives

    J. Marriage Fam.

    (1968)
  • E.R. Dugundji et al.

    Discrete choice with social and spatial network interdependencies: an empirical example using mixed generalized extreme value models with field and panel effects

    Transp. Res. Rec.: J. Transp. Res. Board

    (2005)
  • Ehrhart, B., Mildenhall, S., Jakubowski, S., Podlaha, A., Bowen, S., 2012. Annual Global Climate and Catastrophe...
  • C.S. Fischer

    To Dwell Among Friends: Personal Networks in Town and City

    (1982)
  • Sheep, Floating, 2012. Floating...
  • H. Fu et al.

    Sequential logit dynamic travel demand model for hurricane evacuation

    Transp. Res. Rec.: J. Transp. Res. Board

    (2004)
  • H. Fu et al.

    Survival analysis-based dynamic travel demand models for hurricane evacuation

    Transp. Res. Rec.: J. Transp. Res. Board

    (2006)
  • C.H. Gladwin et al.

    Modeling hurricane evacuation decisions with ethnographic methods

    Int. J. Mass Emerg. Disasters

    (2001)
  • H. Gladwin et al.

    Social science research needs for the hurricane forecast and warning system

    Nat. Hazards Rev.

    (2007)
  • Cited by (47)

    • Mapping the knowledge domain of soft computing applications for emergency evacuation studies: A scientometric analysis and critical review

      2023, Safety Science
      Citation Excerpt :

      In recent years, the increasing frequency of extreme weather and natural disasters, such as wildfires, hurricanes, earthquakes, tsunamis, and floods, has posed devastating threats to human life and social stability (Liu et al., 2021a). Examples include the Wenchuan earthquakes in China in 2008 (Song et al., 2019), Australia's catastrophic 2019/20 bushfires (Filkov et al., 2020), Hurricane Sandy (October 2012) in the United States of America (U.S.) (Sadri et al., 2017), and the 2011 Tohoku Earthquake Tsunami in Japan (Takabatake et al., 2020). In addition to these natural disasters, newly emerging cases of hazards, such as chemical accidents, terrorist attacks, and crowd-stampedes, have led to significant human casualties and property losses as well (Liu et al., 2020).

    • Mechanism of an evacuation cascade during the 2011 Tohoku tsunami inferred from an evacuation simulation incorporating communications in social networks

      2022, International Journal of Disaster Risk Reduction
      Citation Excerpt :

      The results highlight the potential significance of communications within communities that can enhance or hinder evacuations in local communities. As previous observations (e.g. Refs. [8,10,12–14]) have suggested, the results of the model developed in this study clearly show the effects of communication in social networks on enhanced evacuation behaviours (e.g., case 3 in the numerical experiments). However, the model demonstrates that such communications can also hinder evacuation behaviours in communities (e.g., case 2 in the numerical experiments).

    • Activities and social interactions during disaster evacuation

      2021, International Journal of Disaster Risk Reduction
    View all citing articles on Scopus
    View full text