Evaluating drivers of coastal relocation in Hurricane Sandy affected communities

International Journal of Disaster Risk Reduction 13 (2015) 215–228

Contents lists available at ScienceDirect

International Journal of Disaster Risk Reduction journal homepage: www.elsevier.com/locate/ijdrr

Evaluating drivers of coastal relocation in Hurricane Sandy affected communities Anamaria Bukvic a,n, Andrew Smith b, Angang Zhang c a

Urban Affairs and Planning, Virginia Tech, 213 Architecture Annex, Blacksburg, 24061 USA Department of Psychology, Virginia Tech, 109 Williams Hall, Blacksburg, VA 24061, USA c Department of Statistics, Virginia Tech, 406 Hutcheson Hall, Blacksburg, VA 24061, USA b

art ic l e i nf o

a b s t r a c t

Article history: Received 13 April 2015 Received in revised form 18 June 2015 Accepted 18 June 2015 Available online 23 June 2015

The future viability of some coastal communities has been severely challenged by the recent major disasters, as well as other episodic and chronic coastal hazards. These events also instigated a dialogue on their long-term resilience, adaptation options, and the possibility of permanent relocation from high risk areas. Little is known how exposure to disaster, in combination with other contemporary coastal challenges, affects willingness to consider relocation on a household level in the highly-developed urban settlements. The main objective of this paper is to provide a bottom-up perspective on this dilemma via identification of demographic determinants and other disaster-related concerns that may influence support for relocation. More specifically, this study takes an interdisciplinary approach to examine the effects of pre-disaster socio-economic household characteristics, level of preparedness, disaster exposure, experience with recovery, community embeddedness, and resource loss on relocation decisionmaking. The findings hereby reveal that the willingness to consider relocation is primarily influenced by the age of respondents, disaster exposure, level of experienced stress related to recovery, personal financial recovery concerns, future cost of living in high-risk area, concerns with increase in crime and future flooding, and disasterinduced resource loss. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Relocation Retreat Adaptation Coastal Hurricane Sandy

1. Introduction Coastal communities are increasingly exposed to impacts of accelerated climate change such as more intense, longer-lasting, and wetter hurricanes [16,17,36,59,60] and sea-level rise (SLR) [34,46,47,63,29,11,,48]. In combination with persistent coastal problems like erosion and land subsidence [8], the chronic and episodic nature of these events can degrade natural inundation buffers such as barrier islands, dunes, and wetlands, eventually leading to a more frequent and prolonged tidal flooding [33,46,62]. Long histories of unsustainable coastal land use and development patterns that foster high population densities and urban growth further contribute to the overall complexity of coastal issues [45]. The aforementioned conditions can exert a significant stress on social, legal, environmental, and economic sectors in coastal urban areas [3] and cause an extensive damage to infrastructure, public and private property, and productive agricultural land, potentially displacing millions of people [49,45]. In response to emerging n

Corresponding author. Fax: þ 1 540 231 3367. E-mail addresses: [email protected] (A. Bukvic), [email protected] ([email protected] (Zhang). http://dx.doi.org/10.1016/j.ijdrr.2015.06.008 2212-4209/& 2015 Elsevier Ltd. All rights reserved.

climate change impacts and based on the value of structures, adaptation costs, socio-cultural significance, resources, and overall vulnerability, communities have three main options: do nothing, protect themselves, or relocate to a safer location [1]. In the recently published 3rd National Climate Assessment report [21], the authors state, “As sea level raises faster and coastal storms, erosion, and inundation cause more frequent or widespread threats, relocation (also called (un)managed retreat or realignment), while not a new strategy in dynamic coastal environments, may become a more pressing option”. The report further notes that “up to 50% of the areas with high social vulnerability face the prospect of unplanned displacement under the 1–4 foot range of projected sea level rise” due to financial inability to afford structural protection, difficulty to justify public expense, and lack of social and political support for more orderly retreat. Although relocation may represent the most effective long-term adaptation strategy for some coastal communities, this option is still largely considered outside the range of acceptable options due to political, institutional, socio-cultural, and economic considerations. However, a direct exposure to disaster as a discernable and amplified manifestation of other more gradual but chronic hazards in inherently vulnerable coastal locations can serve as a catalyst for a debate focused on questions surrounding relocation vs.

216

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

reconstruction [26]. Even though structural interventions and flood-proofing have been preferred coping strategies among Hurricane Sandy-affected residents [9], the reality is that we simply cannot protect everything [52] and, at some point, will have to consider retreat. This paper explores factors that may affect perceptions and consideration of relocation as a response strategy to contemporary coastal hazards. More specifically, it presents results from a survey targeting households recently exposed to a major hurricane event under the assumption that such experience may heighten people's awareness of coastal risks and advance their thinking of possible solutions. The analysis evaluates a broad range of contextual coastal factors as possible drivers of relocation decision-making (Fig. 1), from the household level socio-economic indicators to various other disaster-related stressors like psychosocial and physical impacts, post-disaster recovery concerns, and relocation assistance support needs. Disasters like Sandy offer a unique window of opportunity for the reexamination of community capacity to withstand episodic and chronic hazards, for the adjustment of policy and planning frameworks to better match the risks, and for the effective change of land use patterns to move critical facilities, assets, and people out of the harm’s way. The results presented herein provide an indication of varying preferences and concerns that drive consideration to relocate among disaster-affected coastal residents and as such provide the vital information on circumstances that may generate a greater support for this adaptation option. Overall, the literature demonstrates that socio-economic factors play an important role in risk perceptions and relocationrelated decision-making. For example, [5] found that natural disaster-induced displacement is influenced by factors such as race/ethnicity, wealth, homeownership, education, age, and gender. Landry et al. [38] also established that the return migration of Hurricane Katrina evacuees was affected by household income, age, education level, employment, marital and homeownership status, albeit with some variation in responses among different population groups. Thus, the measurement of socio-economic characteristics can serve as a useful predictor of willingness to relocate, as it relates to the concept of social vulnerability developed in the context of natural hazards and disasters. According to Cutter et al. [14] and Adger [2], such characteristics may

Fig. 1. Conceptual framework of research design.

modulate individual and community responses to disaster and include, for example, age, gender, ethnicity, employment, and affluence. They can also reflect varying ability of people to engage in preparedness, response, and recovery from different hazards [18], potentially affecting willingness to consider relocation. To account for the importance of diverse range of personal and situational factors in adaptation and disaster risk reduction decision-making, this study evaluates a comprehensive portfolio of contemporary contextual considerations as potential relocation drivers, such as extent and duration of disaster exposure, socio-economic circumstances, post-disaster community disruption and satisfaction with recovery process, as well as risk perceptions. Disasters frequently reintroduce dilemma whether to rebuild in high-risk locations or relocate, both among the officials and affected residents. This issue whether to return from evacuation or stay in host community has been previously explored in the context of coastal disasters – Hurricanes Katrina, Andrew, and Sandy [15,24,38,57]. However, it likely differs from the anticipatory decision-making on relocation that should preferably take place under non-emergency circumstances when the immediate sustenance needs, priorities, and concerns are addressed. The need for inclusion of relocation in the portfolio of climate change adaptation strategies has been increasingly recognized by decision-makers and other stakeholders. This is evidenced by, for example, establishment of acquisition or buyout programs post Hurricane Sandy aimed at purchasing damaged properties located in high-risk areas. The New York State Governor Andrew Cuomo established a buyout program to incentivize relocation by compensating participating homeowners the pre-storm market value of their home and offering additional 5 percent bonus to those who stay locally or 10 percent to groups who sell collectively [32]. Generally, these programs have very low participation rates [6], perhaps due to their simplistic design solely based on financial exchange with disregard of the diverse household contexts and needs. Another reason for limited participation in buyouts may be the unsuccessful effort to effectively engage stakeholders in participatory dialogue on the actual risks and realistic response options. Understanding the complexity and needs associated with contemporary coastal population movement requires an integrative, transdisciplinary approach that builds upon existing research on migration and displacement at different levels of analysis [10,22,39,4,41,53,7] and also accounts for the new emerging aspects of this issue. Even though research on environmental displacement, resettlement, and migration is well established, it often depicts partially or fully reversible conditions and still rarely reflects the scope of environmental changes that are likely to occur due to accelerated climate change. According to Kniveton [35], a number of predicted climate change outcomes are likely to be of a magnitude and variability rarely experienced by the communities in the past, further limiting the applicability of existing statistical models, scenarios, and historical analogs to extrapolate future population shifts induced by climate change. Considering individual households likely have dissimilar needs and preferences for relocation assistance, the ideal relocation programs would include flexible, incentivized, and customized features, rather than standardized assistance packages, strict deadlines, and equal participation requirements. Webber [4] states that the fewer choices people have for moving, it is more likely that the outcomes of that movement will be negative. Strategies that allow people to select between diverse choices, such as solely financial assistance/compensation, social services, alternative housing, new employment, or any combination of these, may represent a more appealing incentive to potential relocatees. To explore which of these incentives in relocation programs are the most useful to coastal residents, this survey includes a few items proposing different

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

support options. The psychological dimensions of disaster experience may also contribute to willingness to relocate and were considered in this study. For example, the resource loss measures the extent to which individuals lost personally and objectively valuable resources as a result of the disaster and its impact on their overall psychosocial wellbeing [28]. Such losses may play a role in the extent to which individuals feel grounded in and committed to “staying put” in the process of recovering from disasters.

2. Data collection and analysis This paper examines the Hurricane Sandy disaster as a case study that demonstrates the full complexity of multidimensional challenges, vulnerabilities, and adaptation needs of urban coastal communities at high risk of future impacts. This event led to extensive housing loss and property damage [20], disruption in transportation and commercial operations, lack of electricity, shortage of food, medicine, and gasoline, as well as population displacement [61]. Widespread impacts across multiple sectors and systems experienced after Sandy reiterate the importance of holistic assessment of the complex interplay between geophysical risks and various socio-economic, cultural, and institutional community characteristics in order to identify the exact vulnerabilities of place and the most effective response strategies [56].

217

2.1. Survey strategy Data for the current study were collected through door-to-door surveys among Hurricane Sandy-affected communities five months after the disaster (May 28–June 05, 2013). The survey sites were identified using FEMA's Remotely-Sensed Damage Assessment data to generate a map of Hurricane Sandy high-impact areas along the U.S. Northeast coast. Within this range, the following NJ/NY communities sustaining the highest levels of structural damages were selected as study locations: Ventnor City, Longport, Margate City, Lavallette, Pine Beach, Manasquan, Belmar, and Long Beach (Fig. 2). The survey was conducted in the randomly selected high-density residential neighborhoods with single-family homes within a few neighborhood blocks or 1-mile radius from the shoreline. A total of 125 surveys were collected during the daylight hours with each session typically lasting 15– 20 min. Any permanent homeowner of a single-family residence willing to complete the survey (IRB #11-725) participated in the study. This analysis is based on 118 responses with 5.6% missing variables. The questionnaire measured the following demographic variables: gender (male, female); age (year of birth); ethnicity (African-American, Asian, Hispanic, Native American, Pacific Islander, White); highest level of education (less than high school, highs school, certificate or associate degree, college/bachelor degree, graduate school, doctorate); employment (full-time employed, part-time employed, self-employed, unemployed, stay-at-home parent, retired, prefer not to answer); income ($1–14,999,

Fig. 2. Study locations in Hurricane Sandy high impact area (with high impacts counties in dark gray and high impacts counties in light grey; source: FEMA's MOTF’s Hurricane Sandy analysis).

218

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

$15,000–29,999, $30,000–49,999, $50,000–74,999, $75,000– 100,000, $100,000 and above); and length of residence in the household ( o1 year, 1–3 years, 3–5 years, 5–10 years, 10–20 years, 20–30 years, 430 years). Sample demographics consisted of all white, 50% female respondents, and 90% age 45 and older. The majority have college degrees (n ¼43), followed by individuals with graduate degree (n¼ 30), high school degree (n¼ 29), associate degree (n ¼19), less than high school education (n ¼ 2), and doctorate degree (n¼ 2). Forty-seven percent of respondents were retired, 28% full-time and 11% part-time employed, 7% self-employed, 4% stay-home parent, and 2% unemployed. The surveyed households predominantly belong to a higher income bracket with 47% earning more than $100,000, 18% $75,000–99,999, 18% $50,000–74,999, 10% $30,000–49,999, and 6% less than that. A vast majority of people were long term residents with 26% residing in the same house for more than 30 years, 18% 20–30 years, 29%10– 20 years, 14% 5–10 years, 6% 3–5 years, 6% 1–3 year, and 1% less than a year. Additional items were documented to indicate the level of preparedness (ownership of flood and property insurance before Hurricane Sandy: yes or no), exposure (damage to house and property: no real damage, mild-broken trees and yard debris, mild to moderate-broken trees, siding damage, and some flooding, moderate-damage to roofing, doors and windows, ground level flooding, moderate to severe-structural house damage and severe flooding, severe-roof failure, wall collapse, and major flooding, and length of time in displacement before being able to return to house: immediately (1–2 days), 1 week, 2–4 weeks, 2–3 months, 3–6 months), and community embeddedness (I work in the community, I have family in the community, I have friends in the community, I have the same ethnic group in community, I have favorite amenities in the community, I like the vicinity of the beach and ocean, I like the layout of the community). A majority of residents were able to return shortly after the Sandy's impact with 28% returning immediately or never left and 29% doing so within 1 week. However, 43% of residents were displaced for extended period of time, with 7% away for 2–4 weeks, 19% for 2–3 months, and 17% for 3–6 months. While only 66% of households had flood insurance and 34% did not, this number was higher for personal property protection insurance with 86% holding coverage and 14% lacking one. More than half of the households experienced moderate to severe home and property damage, with 34% moderate extent (damage to roofing, doors and windows, ground level flooding), 22% moderate to severe (structural house damage and severe flooding), and 5% severe (roof failure, wall collapse, and major flooding). Only 20% of households did not experience any damage, while 11% had mild loss (broken trees and yard debris) and 7% mild to moderate (broken trees, siding damage, and some flooding). Finally, we measured respondents' attitudes and perceptions towards the broad range of factors relevant to relocation decisionmaking process. Their selection was largely inspired by the postSandy media reports and informal conversations with affected residents and local stakeholders, as well as from the similar survey conducted in Sandy affected area a few months earlier than this project [9]. Despite the best effort to include a comprehensive portfolio of conditions and influences that may drive coastal relocation, the authors recognize that variables affecting this process likely include additional location- and household-specific factors. This survey section included three questions exploring relocation potential or pre- and post- disaster community factors potentially impactful on decision to consider relocation. The measures were constructed by the first author of the current paper in 2013 and offer a portfolio of community or personal level factors related to residents' relationship with community, recovery concerns, stress induced by thinking about risks and recovery, disaster-induced

secondary/cascading impacts, and preference for the type of relocation assistance support. In total, the three sets of items were utilized to examine relocation potential via a five-level Likert scale (strongly agree to strongly disagree): (1) Disaster-related stress (8 items), (2) Recovery concerns (11 items), and (3) Relocation drivers (10 items). The Resources Loss (RL) was measured using 12 items from the Conservation of Resources Evaluation (COR-E, [27]. Items for the current study were selected from the pool of COR-E questions based on the specific losses associated with Hurricane Sandy. Items were answered on a five-point rating scale (1 ¼no loss to 5 ¼extreme amount of loss). For the current study, all 12 items were aggregated into a total score after demonstrating adequate internal reliability. 2.2. Statistical analysis Analysis of individual questions (relocation potential measures) was performed using Wilcoxon signed-rank test to calculate the mean scores (with higher scores reflecting higher preference for certain answers) and Kruskal–Wallis (KW) test to determine the uniformity of individual responses within measures. Correlation analysis examining the relationship between demographic and other independent variables included Spearman (for ordinal categorical survey responses) and chi-square (for nominal categorical variables) calculation. The phi-coefficient was calculated as a measure of effect size between variables including the socio-economic question with nominal outcomes (presented as contingency tables). The effect size categories proposed in [13] were applied to find pairs of variables that exhibited small (phi ¼0.1), medium (phi ¼0.3), or large effects (phi ¼0.5). Next, we examined linear combinations of the demographic variables and disaster related predictors that were significantly correlated to the primary relocation instruments through simultaneous regressions. Specifically, based on the significant correlations, (a) stress measure was regressed on age, extent of damage, resource loss, and time of return simultaneously; (b) relocation concerns were regressed on damage, return, flood insurance and property insurance simultaneously, and; (c) relocation drivers were regressed on longevity, age, education, and resource loss simultaneously. Following the results of the simultaneous regressions, conditional effects (i.e., moderated effects) of age on stress and relocation concerns were tested using [25] customizable Process software for SPSS 21.0 (Hayes' Model 1). The conditional effects were estimated through repeated bootstrapping resampling (1000 repeated re-samples) to most accurately estimate the confidence intervals and manage measurement error.

3. Results and discussion 3.1. Concerns driving relocation decision-making The first objective of this paper is to identify which contemporary coastal post-disaster concerns are more likely to prompt oceanfront households to consider relocation. Herein, we assume that households exposed to a disaster may have heightened awareness of their episodic and gradual coastal hazards and more carefully evaluate their adaptation and disaster risk reduction options. The statistical analysis of stress, recovery concerns, and relocation drivers' measures confirmed that they are all significantly different from each other and between the individual responses. Table 1 reveals different preferences for responses offered for question ‘What is causing you to feel more stressed in the aftermath of Hurricane Sandy?’ Post-disaster stress stemming from the acute exposure to disaster event, displacement, loss, and recovery, likely plays an

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

important role in risk perceptions, especially those that influence decision to return to affected area and commit to rebuilding and recovery or to relocate. In this case, respondents were primarily stressed with rebuilding and recovery after Hurricane Sandy, followed by the threat of recurrent hazards, and then with filing of insurance claims. Other items of lesser importance include loss of personal belongings, mold and corrosion, future in this community, and looting and crime. The same or similar concerns were also reported in the media and personal conversations with affected stakeholders, as well as directly observed by researchers during the data collection visit to Hurricane Sandy affected communities five months after the event (Fig. 3). Moving somewhere is perceived as the least stressing factor post-Hurricane Sandy. This may indicate that respondents consider this option less disturbing than dealing with the future risks and recovery challenges. On the other hand, it may also mean that they successfully returned, reestablished their livelihoods, and do not consider relocation, thus eliminating any potential stress associated with this option. Table 2 shows ranked responses for the recovery concerns measure, guided with question ‘You decided to return. Which of these concerns would prompt you to consider relocation in the future?’ The majority of respondents selected the increase in insurance rate, followed by the tax increase and then tidal inundation/frequent flooding, as their primary concerns. This finding shows a substantial concern with financial instruments designed to indicate the level of risk and the actual cost of living in hazardprone area. It also suggests that the increase in premiums may represent an effective tool for coastal land use management. Even though some proponents call for a more aggressive and widespread use of economic disincentives to guide urban growth in coastal communities, their implementation is facing a number of barriers such as political and public opposition, issues with enforcement and enrollment, and social injustice [37]. Some of these issues can be circumvented by provision of subsidies, vouchers,

219

rebates, and community development grants for the low-income homeowner's and small businesses, as well as by active participation and engagement of coastal stakeholders in this issue [12]. The respondents' high-ranking concern for tidal inundation/ frequent flooding and recurrent hazards like nor’easters and hurricanes, demonstrates their heightened awareness and recognition of both acute and chronic nature of these impacts. The concerns with the crime increase, new FEMA advisory maps, city rebuilding requirements, uncertainty when flooding will occur, neighbors, friends and family moving out, strangers in the neighborhood, and construction crews and activities had a lesser impact on the willingness to consider relocation. The last three lowest-ranking

Table 2 Response preferences for the recovery concerns measure (Wilcoxon Signed Rank Sums test). Concerns prompting consideration of relocation in the future 1 2 3 4 5 6 7 8 9 10

Insurance rate increase Tax increase Tidal inundation and frequent flooding Crime increase New FEMA advisory maps City rebuilding requirements Uncertainty when flooding will occur Neighbors, friends, family move out Strangers in the neighborhood Construction crews and activities

Table 3 Response preferences for the relocation drivers measure (Wilcoxon Signed Rank Sums test). Would consider relocation in the future if

Table 1 Response preferences for the stress measure (Wilcoxon Signed Rank Sums test). Causes of stress in the aftermath of Hurricane Sandy 1 2 3 4 5 6 7 8

Rebuilding and recovery Recurrent hazards Filing insurance claims Loss of personal belongings Mold and corrosion Future in this community Looting and crime Moving somewhere else

1 2 3 4 5 6 7 8 9 10 11 12 13

Crime becomes worse We have two or more floods in the next few years Insurance cannot cover full reconstruction Services and amenities do not restore their full function We have one more flood in the next few years School system deteriorates I am offered financial compensation (buyout) Businesses do not reopen I am offered with comparable housing in similar community elsewhere Neighbors, friends, and family move out I am provided with free legal service I receive assistance with finding a new job elsewhere I can move together with my neighbors

Fig. 3. Lingering Hurricane Sandy impacts in the affected communities six months after the event.

220

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

concerns exploring the importance of social aspects and disruption in the community social fabric on willingness to consider relocation did not represent a significant consideration for this inquiry. Table 3 shows response preferences for the similar question as before (whether respondents would consider relocation in the future), albeit offering a different set of answers depicting more personal/household level stressors and a few of potential relocation assistance support options. Here, the majority of responders would consider relocation if the crime becomes worse, if they experience two or more floods in the next few years, and the insurance cannot cover full reconstruction. They would be less likely to relocate if services and amenities do not restore their full function, if they experience one more flood in the next few years, school system deteriorates, they are offered a buyout program, businesses do not reopen, etc. The option to move together with their neighbors, collectively as a group, was the least important reason to consider relocation. This suggests that households in surveyed locations prefer to make decisions related to relocation independently from the rest of their community. The shortcoming of this tendency is that it may result a “Swiss cheese” pattern of the community with randomly distributed vacant lots that can further undermine the neighborhood viability and resilience. Similar outcome was observed following the buyout of 9000 properties in Iowa after 1993 Midwest floods, with a limited gain in flood risk reduction and ecologically sound open space [23]. To address this concern with checkerboard participation, the New York Governor Cuomo proposed post-Sandy an innovative buyout program offering10% bonus to all households who agree to move as a whole block [55]. Moreover, the random retreat pattern may also drive the resentment among remaining residents due to feeling of social

abandonment, negative impact on the real estate values, increase in crime, inefficient delivery of services and infrastructure, and discouragement of investment in the community. The authors hereby recognize that relocation potential likely varies in different geographic and socio-cultural contexts and that some communities may prefer to relocate together to preserve social networks and cultural, ethnical, linguistic, and other types of connectivity. Therefore, highly homogenous and cohesive communities such as villages Newtok and Shishmaref in Alaska or Smith Island in Chesapeake Bay likely have very different perceptions and preferences as of the character of this movement. In addition to socio-economic characteristics, the survey explored the role of community embeddedness to capture respondents' relationship with the community and how this may influence decision to consider relocation. A majority of respondents (78%) like the vicinity of beach and ocean, 73% have friends in the community, 73% like the community layout, 71% have favorite amenities in the community, 59% have same ethnic group in the community, almost half (46%) have family in the community, and only 24% work in the community. The data suggests that surveyed group has well-established social ties within the community and strong affinity for the oceanfront living, but much lesser need to reside there for the economic sustenance. 3.2. Impact of socio-economic determinants on relocation potential The overall relationship between demographic and disaster exposure/preparedness variables and measures of relocation potential was further explored to determine the extent and type of their interaction (Table 4; also Appendix, Tables A1–A6). The

Table 4 Summary of correlation analysis results (available in Appendix) between socio-economic variables and relocation potential measures. Category Gender

Group Males

Moew STRESSED with

More CONCERNED with Insurance rate increase

More likely to RELOCATE if

Age

Younger

All offered factors

Crime becomes worse School system deteriorates Services and amenities do not restore Insurance cannot cover full reconstruction Receive assistance with finding new job

Employment

Fulltime

Receive assistance with finding new job

Education

Higher

Rebuilding and recovery Future in community Lost personal belongings Filing insurance claims Mold and corrosion

Tax increase Insurance rates increase New FEMA advisory maps City rebuilding rules Strangers in the neighborhood Crime increase Strangers in the neighborhood

Strangers in neighborhood Crime increase Tidal inundation/frequent flooding

Experience two or more floods Crime becomes worse Services/amenities do not restore function

Income Flood Insurance

Higher Yes

Experience two or more floods

Property Insurance

Yes

Length of Residence

Shorter

Rebuilding and recovery Filing insurance claims

New FEMA advisory maps Crime increase Neighbors, friends, family moving out Construction crew and activities Tidal inundation/ frequent flooding

How Soon Returned

Later

Experienced Damage

More

Recurring hazards Rebuilding and recovery Future in community Mold and corrosion Lost personal belongings Recurring hazards Rebuilding and recovery Filing insurance claims Future in community Mold and corrosion Lost personal belongings

Insurance rate increase

New FEMA advisory maps City rebuilding rules Tidal inundation/frequent flooding

Receive assistance with finding new job Receive legal advice on options Experience one more flood Experience two or more floods Neighbors, friends, family move out Businesses do not reopen Crime becomes worse School system deteriorates Services/amenities do not restore function Neighbors, friends, family move out

Experience one more flood Experience two or more floods Neighbors, friends, family move out Crime becomes worse School system deteriorates

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

correlation between stress measure and demographic profile reveals that older respondents are significantly less stressed with all proposed considerations than the younger participants. More educated respondents are more stressed when thinking about filing insurance/assistance claims and mold and corrosion, indicating possible frustration with the administrative hurdles and financial considerations of recovery, as well as greater awareness of secondary flood damages that can impede renovation efforts and adversely affect their health and property value. Income does not represent a significant determinant of level of stress experienced when thinking about post-disaster concerns. Respondents who lived in surveyed households for shorter time were more stressed when thinking about rebuilding/recovery and filing insurance/assistance claims. Those who were displaced by Sandy for the prolonged period were more stressed with almost all offered considerations, including thinking about recurrent hazards in coastal areas, rebuilding and recovery, future in this community, mold and corrosion, and lost personal belongings. Likewise, respondents who experienced more property damage selected all the items as afore with the addition of concern with filing insurance/assistance claims (Table A1). Table A2 shows correlation between responses capturing general recovery concerns that may affect willingness to consider relocation in the future and socioeconomic variables. The older residents are less likely to consider relocation due to financial and bureaucratic considerations related to living in high risk location such as higher taxes and insurance premiums, new FEMA advisory maps, and city rebuilding rules, as well as safety concerns – strangers in the neighborhood and crime increase. More educated respondents would consider relocation in the future if they would observe more strangers in the neighborhood, crime increase, and increase in tidal inundation and flooding. Again, the income does not represent a significant determinant in the selection of responses within this measure. Respondents who lived for a shorter period of time in the household are more likely to support relocation due to concerns with tidal inundation and frequent flooding, perhaps because they are still not fully assimilated into the community and have less emotional bias in risk assessment. Respondents who returned home after the extended period of time post Hurricane Sandy perceive flood insurance rate increase as the greatest concern that would prompt them to relocate in the future. Those who self-reported higher levels of experienced property damage were significantly more concerned with the new FEMA advisory maps and city rebuilding rules, as well as with tidal inundation and frequent flooding, indicating increased awareness of long-term risks facing coastal communities and ensuing regulatory changes that may affect reconstruction efforts. This finding supports the previous research showing a positive relationship between experienced hurricane/flood damage and level of risk perception [19,40,50,54], as well as an important impact of personal experience with hazard event on the risk perception of future hazards discussed in [42]. Next, we evaluated the relationship between relocation drivers measure and socioeconomic and disaster exposure variables (Table A3). Younger people are more likely to relocate when the crime becomes worse, school system deteriorates, services and amenities do not restore their full function, insurance cannot cover full reconstruction, and they are offered with assistance in finding a new job elsewhere. The educated respondents are more likely to consider relocation in the future should they experience two or more floods in in next few years, crime becomes worse, and services and amenities do not restore their full function. Income once again did not represent an important factor in response selection. Respondents who lived in the surveyed house for a shorter period of time would relocate should the community experience one and two or more floods in the next few years, neighbors friends and

221

family move out, businesses do not reopen, crime becomes worse, school system deteriorates and services and amenities do not restore their full function. This latest finding may suggests that newer residents moved into this community specifically for the very same reasons they listed as potential concerns that would prompt them to relocate. When these pull forces that attracted them to move into the community would cease to exist, they would lose the incentive to stay and cope with coastal hazards. Participants who returned back after a prolonged time in displacement would consider relocation if their neighbors, friends, and family move out. Those who experienced extensive damages would relocate if they are facing any additional flooding in next few years; neighbors, friends and family move out; crime becomes worse; and school system deteriorates. Only respondents who experienced significant disaster impacts, observed as a direct property damage and prolonged time in displacement, and also lived in community for a long time, were concerned whether their neighbors, friends, and family decide to move out. Additional demographic factors evaluated as a part of this study include gender, employment, ownership of flood insurance and personal property insurance (Appendix, Tables A4–A6). Gender has the minimal effect on thinking about repetitive flooding and filing insurance/assistance claims and overall does not serve as an important indicator of disaster-related stress (Table A4). The only large effect was observed for the employment category with the full time employed respondents being more stressed when thinking about rebuilding and recovery, future in this community, and lost personal belongings. The respondents with flood insurance were least stressed when thinking about rebuilding and recovery and those with personal property insurance were least stressed when thinking about recurrent hazards, mold and corrosion, lost personal belongings, recurrent damages, and the overall future in this community. Gender has significant impact only on concern with the insurance rate increase with male respondents being significantly more concerned with this option (Table A5). It also has the statistically smallest effect on concerns with new FEMA maps, city rebuilding rules, strangers in the neighborhood, and uncertainty when the next flooding will occur. Employment has medium to large effect on concern with strangers in the neighborhood as a driver of relocation with fulltime employees being the more concerned with the strangers in the neighborhood when compared to unemployed, stay-at-home moms, and retired people. The lower concern with the presence of strangers in the area among this latter subgroup may suggest their increased confidence in the whereabouts and identity of outsiders due to their ability to observe daytime activities in their neighborhood. People who own flood insurance are significantly more concerned with new FEMA advisory maps and crime increase, and those who own personal property insurance are significantly more concerned with neighbors, friends, and/or family moving out and construction crews and activities. For the relocation drivers measure (Table A6), the only large effect was observed between employment and relocation support option to receive assistance with finding a new job elsewhere. Other statistically significant values were observed between flood insurance ownership and risk of two or more floods in the next few years and personal insurance and preference for assistance with finding alternative employment elsewhere and provision of free legal advice on response options. The proposed indicators of community embeddedness were explored and subdivided into two categories reflecting actual (work/have family/have friends in the community) and superficial (I have favorite amenities in the community, I like vicinity to beach and ocean, I like layout of the community) perceptions of benefits associated with living in coastal community. Next, the Maslow’s hierarchy of needs theory [44] was tested by measuring whether

222

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

the effect of community embeddedness on recovery concerns and relocation drivers varies as a function of higher resource loss. This concept is represented in a shape of pyramid with the largest area devoted to the most fundamental needs e.g. air, water, food, and shelter. It further suggests that when these basic physiological necessities are fulfilled, people direct their attention to others needs in the successive order: physical and economic safety and security; love and belonging; esteem, and lastly self-actualization [44]. The results show that at low levels of resource loss, the superficial community embeddedness has a direct positive influence on recovery concerns, but at a higher levels of resource loss this effect no longer has effect on this measure. This suggests that when resource loss is high and the surivival and importance of basic needs at the foundation of Maslows hierarchy are paramount, superficial concerns (e.g., esthetics of buildings, landscape, etc.) are no longer salient. It also indicates that personal valuebased preferences for living next to the shore can alleviate risk concerns only up to a certain threshold or tolerance level of exposure, after which they become trivial and household shift their concerns to factors affecting their fundamental needs and wellbeing. 3.3. Predicting relocation variables and disaster-related stress In order to examine predictors of relocation variables (recovery concerns and relocation drivers) and disaster-related stress, we first examined bivariate correlations among (a) the demographic and socioeconomic variables (age, education, gender, and income), and (b) disaster related predictors (damage, resource loss, flood insurance, personal insurance, how soon returned after the hurricane, longevity in neighborhood, community embeddedness [working in community, and having friends/family in the community], and superficial community embeddedness [I have favorite amenities in the community, I like the vicinity of the beach and ocean, I like the layout of the community]) with primary relocation potential instruments. Examination of these correlations demonstrated that stress measure is significantly related to age (r ¼  0.347, p o0.001), the extent of property damage (r ¼0.254, p ¼0.005), how soon respondents returned to their homes since evacuation (r ¼0.301, p ¼0.001), and resource loss (r ¼0.458, p ¼0.001). The other factors like gender, education, income, ownership of flood/property insurance, and how long they resided in the surveyed household are not significantly related to this variable. The recovery concerns are significantly related to the extent of property damage (r ¼  0.233, p¼ 0.010), flood insurance (r ¼0.193, p ¼0.032), and personal insurance ownership (r ¼  0.212, p ¼0.018), how soon they returned home (r ¼  0.188, p ¼0.044), and actual community embeddedness (r ¼0.425, p ¼0.000). Resource loss was not significantly correlated with this measure, as well as variables age, education, income, for how long they lived in the same house and superficial community embeddedness. Respondents who experienced more property damage, had to stay in temporary displacement for longer period of time, did not have personal property insurance, and experienced more stress with disaster recovery were overall more concerned with items listed within this measure. Those who had flood insurance and strong community embeddedness (work/family/ friends in the community) were overall less concerned with stressors listed in this question. Lastly, the relocation drivers measure was significantly related to length of residence in the household (r ¼  0.183, p ¼0.041), age (r ¼  0.379, p o0.001), education (r ¼ 0.199, p ¼0.026), and resource loss (r¼ 0.343, p o0.001). Those residents who lived for a long time in the same house are less concerned with offered responses (r¼  0.183, p ¼0.041), while those who experienced higher levels of stress are significantly more concerned and willing

to consider relocation (r ¼0.663, p¼ 0.000). Older people are less concerned with items listed in relocation drivers measure, while more educated people show the opposite trend. The respondents who experienced more resource loss also suffered more disasterinduced damages (r ¼ 0.510, p ¼0.000), were in displacement for extended period of time (r ¼0.432, p¼ 0.000), felt more stressed with recovery (r ¼0.458, p ¼ 0.000) and were more likely to consider relocation based on the criteria offered in the relocation drivers measure(r ¼0.343, p ¼0.000). Those owning flood insurance were less worried about recovery concerns (r¼ 0.193, p¼ 0.032), while those with personal property insurance were overall more concerned with the same problems (r ¼  .212, p¼ 0.018). Overall, income and gender do not represent an important determinant in majority of responses. Following correlation analysis, we tested linear combinations of the demographic variables and disaster related predictors that were significantly correlated to the primary relocation potential instruments through simultaneous regressions. Specifically, based on significant correlations, a) stress measure was regressed on age, property damage, resource loss, and how soon they returned home after the disaster simultaneously; b) recovery concerns were regressed on age, education, for how long they lived in the household, and resource loss simultaneously, and; (c) relocation drivers measure was regressed on the extent of damage, how soon returned, flood insurance and property insurance simultaneously. Findings show that the model predicting disaster-related stress (including age, damage, when respondents returned to their home, resource loss [each of these variables have an initial significant bivariate correlation with disaster related stress]) accounted for 26% of the variance (adj. R2 ¼ .256, F[4, 110] ¼10.47, po 0.001). Only age (B ¼  0.16, p¼ 0.001) and resource loss (B ¼0.32, p o0.001) predict disaster related stress, where being older predicted lower stress, and higher levels of resource loss predicted higher stress levels. Regression findings predicting recovery concerns (via variables identified as significant in the bivariate correlations in relation to this measure]) revealed that 11% of the variance in this question is explained by this model (adj R2 ¼0.11, F[4, 111] ¼ 4.42, p ¼0.002). Among the independent variables entered into the simultaneous regression, only flood insurance was a significant predictor of recovery concerns (B¼ 4.98, p¼ 0.029). Regression findings predicting relocation drivers demonstrated that the model (including resource loss, for how long respondents lived in the household, education, and age [as the only variables significantly correlated with this measure in the bivariate correlation analysis]) accounted for 23% of the variance in this question (adj R2 ¼ 0.23, F[4, 122] ¼ 10.22, po 0.001). Findings further revealed that among the independent variables entered simultaneously in this model, only age (B ¼  0.29, p o0.001) and resource loss (B ¼0.42, p o0.001) significantly predicted relocation drivers measure. 3.4. Conditional effects predicting stress and relocation drivers measures Considering the simultaneous regression findings that higher age reduced disaster related stress, and that higher resource loss increased the stress levels, we tested a model examining whether the effect of age on distress was moderated by resource loss. Findings show that the model which included age, resource loss, and the interaction term (i.e., age  resource loss) accounted for 34% of the variance in the disaster related stress severity (R2 ¼ 0.34, F[3, 119] ¼ 20.47, p o0.001). Age (B¼  0.42, p ¼0.001) and the interaction term (B ¼0.01, p ¼0.009) had significant direct effects on disaster related stress. Resource loss does not significantly directly predict stress when the interaction term is included in the model. Moderation probing showed that as resource loss increased, the

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

buffering effect of age on disaster related distress was reduced until this protective effect was no longer significant at high levels of resource loss. Specifically, at low levels of resource loss (9.68), the direct, stress reducing effect of older age was at its strongest (B ¼  0.28, 95% CI [ 0.400,  0.165]), whereas the protective effect of age was comparatively diminished as level of resource loss rose. Specifically, at mean level of resource loss [19.21], the effect of age on stress was lower than at low resource loss [B¼  0.145, 95% CI [  0.233,  0.058]), and at high resource loss [28.74], the effect of age on stress was reduced to non-significance (B ¼  0.008, 95% CI [  0.158, 0.143]). Based on findings from the simultaneous regression, age and resource loss were again examined in dynamic fashion as predictors of relocation drivers, such that a model was tested examining whether buffering/protective effect of older age on this measure was moderated by resource loss severity. Findings demonstrate that the model which included age, resource loss, and the interaction term (age  resource loss) accounted for 26% variance in this question (R2 ¼0.26, F[3, 119] ¼ 13.72, p o0.001). In this model, age retained its buffering effect on relocation drivers (B ¼  0.583, p o0.001), whereas resource loss (B ¼  0.569, p ¼0.321) and the interaction term (B¼ 0.016, p ¼.080) did not have direct effects on the outcome. Considering the near significant effect of the interaction on relocation drivers, the moderation effects were probed and examined nonetheless. Findings demonstrated a similar pattern to the previously reported moderation findings: as resource loss increased, so too decreased the buffering effect of older age on relocation drivers until at high levels of resource loss age no longer had an effect on response preferences: (a) at low resource loss, protective effect of age on relocation drivers was strongest (B ¼  0.429, 95% CI [  0.623, 0.235]); (b) at mean resource loss, protective effect of age on relocation drivers was comparatively diminished (B ¼  0.278, 95% CI [  0.423,  0.133]); (c) at high resource loss, protective effect of age on relocation drivers was non-significant (B ¼  0.127, 95% CI [  0.375, 0.122]). Lastly, the dynamic models predicting relocation drivers were tested. Each model posed age as the primary independent variable, stress as the primary mediator, and resource loss as a moderator of the indirect effect of age on relocation drivers through disaster related stress. Findings of the conditional indirect effects model predicting relocation drivers demonstrate that the model accounted for 46% variance in relocation drivers (R2 ¼0.46, F[2, 120] ¼51.47, po 0.001). Age retained a significant direct, buffering effect on relocation drivers even when including the other predictors in the model (B ¼ 0.152, p ¼0.019). Conditional indirect effects (moderated-mediation effects) demonstrated that as resource loss increased in severity, the indirect buffering effect of age on relocation drivers through disaster-related stress severity decreased: (a) at low resource loss, the indirect effect of age on relocation drivers through stress severity (B ¼  0.264, 95% CI [  0.400,  0.145]); (b) at mean resource loss levels, the indirect effect of age on relocation drivers through stress severity (B ¼  0.136, 95% CI [  0.238, 0.063]); (c) at high resource loss levels, the indirect effect of age on relocation drivers through stress severity was non-significant (B ¼  .007, 95% CI [ 0.147, 0.126]).

4. Limitations and future research needs Even though this research provides an important preliminary indication of priority considerations driving willingness to relocate among the disaster-affected coastal residents, it has a few potential limitations and contextual features that require further clarification. The first limitation is its exclusively quantitative nature

223

and the lack of qualitative component to elucidate the authentic bottom-up concerns, potentially affecting consideration of relocation among the surveyed group. It is, however, important to note that this project is building upon a pilot survey conducted in the Sandy-affected area three months earlier [9] that included, in addition to quantitative metrics, open ended questions whose responses informed the development of instruments used in this survey. In addition, many concerns explored in this project were identified from the media reports and informal conversations with affected residents. Therefore, they do reflect, albeit in a more informal sense, concerns identified by the Sandy-affected residents as the most problematic issues. A more rigorous qualitative evaluation of individual needs and preferences related to the possibility of relocation will undoubtedly yield more considerations to help guide the development of flexible relocation programs offering a more diverse portfolio of options to accommodate such needs. Such programs will likely have to strike a balance between accommodating diverse personal needs and providing fiscally sound and administratively manageable relocation support mechanisms. The second limitation is related to our sample profile which is not as diverse as initially desired and is not capturing a full representation of different socioeconomic strata that can be found in the communities along the U.S. coasts. Our sampling strategy was primarily driven by the amount of experienced damage and the extent of Sandy’s impact, as well as some other practical reasons like accessibility and limited coverage in the aforementioned similar pilot survey. Therefore, it prevalently represents oceanfront properties belonging to middle- and higher-income homeowners. We do recommend that future research explores attitudes toward relocation in more socially vulnerable coastal communities, especially considering that the recent U.S. 3rd National Climate Assessment [51] suggests that “up to 50 percent of coastal areas with high social vulnerability face the prospect of unplanned displacement under the 1–4 foot range of projected sea level rise”. The same report notes that such localities will not be able to afford structural protection; benefits are insufficient mainly because they reflect only financial considerations and not sociocultural and ecological aspects; and the sociopolitical support for planned retreat is prevalently lacking [51]. Relevant future research should employ larger, more diverse population, and longitudinal sampling at different intervals post disaster to measure relationship between the progress with the recovery, risk perceptions, and relocation-related concerns. Different surveying strategies (e.g., phone, web, mail, phone app) will allow for a more extensive and integrated data collection than in the case of door-to-door canvasing, therefore improving generalizability of findings to other post-disaster contexts. Consequently, this approach will allow for a more sophisticated and rigorous causality testing through ordering variables in a manner that incorporates temporal precedence within models. Whereas our sample is not as diverse and large as we would have liked, our power analysis for statistical modeling of direct, indirect, and moderated /mediated effects indicate that our power for conducting the statistical analyses that we did was adequate. Nonetheless, herein utilized regression-based path modeling approach was developed [25] for testing interrelationships and dynamics (moderation; mediation; moderated mediation) through the bootstrapping approach (allowing for testing random subsets of the sample in order to more accurately estimate/average measurement error) with smaller samples. Lastly, it would be useful to obtain the actual information on the willingness to relocate among coastal homeowners living in high risk areas affected by a major disaster or recurrent flooding. In this paper, we evaluated willingness to relocate indirectly by performing the correlation analysis through drivers and predictors

224

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

(with focus on factors that would prompt them to consider relocations). The main reason for this decision was the concern that such direct question would distract respondents from thinking about complex circumstances that would affect their consideration of relocation and introduce bias into their answers, by focusing too much on the actual possibility of this happening rather than on understanding factors that would drive this option in a voluntary and bottom-up fashion.

5. Conclusion The recent major disasters and other progressive chronic hazards have initiated a debate whether the long-term viability of some highly vulnerable coastal communities is fully achievable, under what cost, and with what policy and planning interventions. Relocation has been increasingly discussed as one of the potentially effective adaptation and disaster risk reduction strategies for the epicenters of extreme physical vulnerability. The main contribution of this paper to the dialogue on relocation includes the identification of priority concerns that may drive willingness to consider relocation after exposure to a major disaster event. It also provides an indication how these responses differ between surveyed socioeconomic groups with different levels of community embeddedness and degrees of disaster exposure and preparedness. Analysis reveals that respondents are overall more concerned with specific factors such as rebuilding and recovery, especially from the financial perspective (e.g. tax increase, insurance rate increase, and filing assistance claims to receive financial support), demonstrating that these economic considerations represent prominent drivers of relocation decision-making on the household level. It also suggests that financial cost of living in high risk coastal areas could serve as a valid signal of the level of actual risk and consequently an important factor in decision whether to relocate or stay in place. Another important concern identified by this analysis is the risk of repetitive exposure to recurrent or nuisance flooding – suggesting that residents are cognizant of chronic hazards and the likelihood of their recurrence. This is a promising finding considering that projections of accelerated climate change clearly identify gradual but permanent hazards like tidal inundation and sea level rise as the greatest threat for many low-laying coastal communities. Laczko and Aghazarm [39] state that even though natural disasters receive the most media and policy attention, gradual environmental changes may have a stronger influence on population displacement. Similarly, Martin [43] states that most new policy amendments cater to disaster-related socio-demographic shifts and that the policies designed to address migration induced by slow-onset climate change are so far lacking. Natural disasters may displace large numbers of people within a very brief time frame, while slow-onset climatic events are more likely to cause the less apparent displacement, but lead to relocation of many more people in a more permanent manner [7]. Although the actual relationship between such environmental changes and human migrations is still not fully clear, there are a number of examples in which environment had a dominant role among other contributing factors in mass mobilization of people [31]. In the context of this study, this heightened concern may reflect respondents' exposure to the post-disaster surge of media reports discussing coastal risks and diminish over time, especially if households successfully restore their livelihoods and quality of life soon after. This initial level of risk awareness and engagement with the issue among residents can be maintained with the innovative and effective education and outreach efforts and community engagement activities, such as regular community meetings and updates on change in risk and response efforts. Crime

increase was also one of the leading concerns that would prompt respondents to consider relocation, suggesting a significant concern for personal safety and property protection. Post Sandy event, the Huffington Post reported on “Sandy damaged homes targeted by criminals” in June 2013, “Sandy victims’ homes looted for Thanksgiving” in November 2013, “Sandy relief centers plagued by looters” in January 2014, “Burglaries up 500 percent in the Rockaways” in March 2014, etc. [30]. Tierney et al. [58] cite similar excerpts from the news in the context of Hurricane Katrina, noting that “both media reporting and official discourse following Hurricane Katrina upheld the mythical notion that disasters result in lawlessness and social breakdown” but they “also presented highly oversimplified and distorted characterizations of the human response to the Katrina catastrophe”. Therefore, it is not fully clear whether the concern with crime stems from the respondents’ personal experiences, preconceptions, and first-hand information, or from their perceptions shaped by the apprehension expressed by other residents and media outlets. Considering the importance of this factor in the household decision-making process, this aspect should be further explored in the future research. The surveyed population was the least stressed when thinking about moving elsewhere, marginally concerned with relocation decisions among their neighbors, friends, and family, as well as prevalently reluctant to move together as a group. These preferences likely vary in different geographic and sociocultural contexts and some other communities may prefer a collective action aimed to preserve existing social networks and cultural, ethnical, linguistic, and other types of connectivity. Therefore, researchers should gain a better understanding of how different community typologies influence relocation decision-making process, needs and preferences related to its implementation, and overall risk perception that may affect support for this disaster risk reduction and adaptation strategy. The income and gender were consistently the least relevant demographic factors affecting engagement with relocation issues, while the age was notably important consideration – with older respondents being significantly less stressed and concerned across all categories, thus less likely to support relocation. The latter finding likely reflects their heightened resilience echoing lifetime experiences, a degree of freedom from the everyday pressures like childcare and work stress, as well as the perception that long-term projections of coastal impacts do not apply to them due to their age and relative residual life expectancy. This may be problematic from the emergency response standpoint, considering that elderly also represent a highly vulnerable group that often needs additional assistance during the evacuation and disaster response efforts, as well as shelter-in place option, than the general population. Therefore, the risk communication efforts should acknowledge this age-related dissonance, explore possible reasons for these observed results, and design outreach efforts that will reflect such distinctive perspective. Even though it is not entirely surprising that respondents who experienced significant property damage and were in displacement for prolonged time were overall more stressed and concerned with almost all proposed considerations and more willing to consider relocation, this outcome still provides the empirical evidence that this group may be more likely to participate in relocation programs than those households with a more positive experiences. Age has a buffering or protective effect on the levels of experienced stress and consideration of relocation, such that older age predicts lower stress and reduced interest in relocation. However, this effect is conditional on the level of experienced resource loss. Specifically, as resource loss increased, being older no longer protected respondents from the disaster-induced stress or consideration of relocation. Therefore, it appears that elderly are not fully impervious to the adverse impacts of disasters and recovery challenges but rather have a different threshold level than

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

their younger counterparts. Considering that relocation process is likely to occur in the incremental and sequential stages rather than all at once, paper provides an insight into which constituents may decide and under what specific circumstances to relocate sooner or later. This household-level perspective on the contemporary coastal disaster-related concerns that could serve as relocation drivers can help inform development of more appealing and effective relocation programs that reflect location-specific conditions and preferences and therefore receive more support for implementation. While it is vital that relocation mechanisms are flexible and diverse in the type and character of support, it is important to note that this flexibility may have limitations and that at some level, it will be difficult to fully customize relocation programs to reflect all individual household needs.

Acknowledgments The authors thank Dr. Russell Jones, Kaushal Amatya, and Graham Owen for invaluable assistance with data collection, the Virginia Tech's Laboratory for Statistical Analysis (LISA) and Chris Frank for supporting this collaboration, as well as Tiona Johnson for producing the GIS map of survey locations. They are also grateful to Dr. Karen Roberto for the patronage of this interdisciplinary effort and the Institute for Society, Culture, and Environment (ISCE) at Virginia Tech for financially supporting this study.

Appendix A See Tables A1–A6.

Table A1 Correlation analysis between response preferences of the stress measure and different demographic categories and disaster exposure variables.

p o 0.05, shaded cells.

Table A2 Correlation analysis between response preferences of the recovery concerns measure and different demographic categories and disaster exposure variables.

p o 0.05, shaded cells.

225

226

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

Table A3 Correlation analysis between response preferences of the relocation drivers and different demographic categories and disaster exposure variables.

p o0.05, shaded cells.

Table A4 Correlation between response preferences of the stress measure and different socio-economic categories measured according to statistically significant values.

italic*, po 0.05) and effect size (dark grey, 4 0.3 medium–large effect; light gray 0.1 ophi o 0.3 small–medium effect, white o0.1 less than small effect).

Table A5 Correlation between response preferences of the recovery concerns and different socio-economic categories measured according to statistically significant values.

italic*, p o 0.05) and effect size (dark grey, 40.3 medium–large effect; light gray 0.1o phio 0.3 small–medium effect, white o 0.1 less than small effect).

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

227

Table A6 Correlation between response preferences of the relocation drivers and different socio-economic categories measured according to statistically significant values.

italic* p o 0.05) and effect size (dark grey, 40.3 medium–large effect; light gray 0.1o phi o0.3 small–medium effect, white o 0.1 less than small effect).

References [1] Alaska Center for Climate Assessment and Policy, Decision-making for At-risk Communities in a Changing Climate, University of Alaska, Fairbanks, 2009. [2] W.N. Adger, Vulnerability, Glob. Environ. Change 16 (3) (2006) 268–281. [3] K.S. Alexander, A. Anthony, T.G. Measham, Managed retreat of coastal communities: understanding responses to projected sea level rise, J. Env. Plan. Manag. 55 (4) (2012) 409–433. [4] J. Barnett M. Webber, Accommodating migration to promote adaptation to climate change Commission on Climate Change and Development, A policy brief prepared for the Secretariat of the Swedish Commission on Climate Change and Development and the World Bank Report 2010, 2009. [5] R. Black, D. Kniveton, K. Schmidt-Verkerk, Migration and climate change: towards an integrated assessment of sensitivity, Environ. Plan. – Part A 43 (2) (2011) 431. [6] R. Braidwood, Telephone correspondence,Correspondence, September 26, Office of Emergency Management, Chesapeake, Virginia, 2013. [7] O. Brown, Climate Change and Forced Migration: Observations, Projections and Implications. Human Development Report, UNDP, New York, NY, 2007. [8] CCSP, see: Climate Change Science Program, in: J.G. Titus, K.E. Anderson, D.R. Cahoon, D.B. Gesch, S.K. Gill, B.T. Gutierrez, E.R. Thieler, S.J. Williams (Eds.), Coastal sensitivity to sea-level rise: a focus on the Mid-Atlantic region. Synthesis and Assessment Product 4.1, Report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research 2009 Government Printing Office (Lead authors: J.G. Titus, K.E. Anderson, D.R. Cahoon, D.B. Gesch, S.K. Gill, B.T. Gutierrez, E.R. Thieler, S.J. Williams, Government Printing Office. [9] Author, Co-author, Should we stay or should we go: attitudes towards relocation post-Hurricane Sandy. Disasters, in press. [10] J.R. Campbell, M. Goldsmith, K. Kosh, Community relocation as an option for adaptation to the effects of climate change and climate variability in Pacific Island Countries (PICs), Asia-Pacific Network for Global Change Research, 2005. [11] J.A. Church, N. White, T. Aarup, W.S. Wilson, P.L. Woodworth, C.M. Domingues, K. Lambeck, Understanding global sea levels: past, present and future, Sustain. Sci. 3 (1) (2008) 9–22. [12] R. Cleetus, Overwhelming risk rethinking flood insurance in a world of rising seas, Report by the Union of Concerned Scientists, Cambridge, MA, 2014. [13] J. Cohen, A power primer, in: A.E. Kazdin (Ed.), Methodological Issues and Strategies in Clinical Research., third edition,American Psychological Association, Washington, DC, 2003, pp. 427–436. [14] S.L. Cutter, B.J. Boruff, L.W. Hirley, Social vulnerability to environmental hazards, Soc. Sci. Q. 84 (2003) 2. [15] J.R. Elliott, J. Pais, Race, class, and Hurricane Katrina: social differences in human responses to disaster, Soc. Sci. Res. 35 (2) (2006) 295–321. [16] J.B. Elsner, J.P. Kossin, T.H. Jagger, The increasing intensity of the strongest tropical cyclones, Nature 455 (7209) (2008) 92–95. [17] K. Emanuel, Increasing destructiveness of tropical cyclones over the past 30 years, Nature 436 (7051) (2005) 686–688. [18] C.T. Emrich, S.L. Cutter, Social vulnerability to climate-sensitive hazards in the Southern United States, J. Weather Clim. Soc. 3 (3) (2011) 193–208. [19] C.E. Faupel, S.P. Styles, Disaster education, household preparedness, and stress responses following Hurricane Hugo, Environ. Behav. 25 (1993) 228–249. [20] FEMA, see: Federal Emergency Management Agency, Hurricane Sandy One Year Later, Available via 〈http://www.fema.gov/hurricane-sandy〉, 2012 (cited 16.10.14). [21] GCRP, see: the U.S. Global Change Research Program, 3rd National Climate Assessment Report. Available via 〈http://nca2014.globalchange.gov/〉, 2014 (cited 02.10.14).

[22] F. Gemenne, Migration, a possible adaptation strategy? Inst. Sustain. Dev. Int. Relat. 3 (2010) 1–4. [23] D. Godschalk, T. Beatley, P. Berke, D. Brower, E.J. Kaiser, Natural Hazard Mitigation: Recasting Disaster Policy and Planning, Island Press, Washington DC, 1999. [24] J.A. Groen, A.E. Polivka, Going home after Hurricane Katrina: determinants of return migration and changes in affected areas, Demography 47 (4) (2010) 821–844. [25] A.F. Hayes, An Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-based Approach, Guilford Press, New York, 2013. [26] J. Henry, Return or relocate? An inductive analysis of decision-making in a disaster, Disasters 37 (2) (2013) 293–316. [27] S.E. Hobfoll, The influence of culture, community, and the nest-self in the stress process: advancing conservation of resources theory, Appl. Psychol.: Int. Rev. 50 (2001) 337–421. [28] S.E. Hobfoll, Traumatic stress: a theory based on rapid loss of resources, Anxiety Res. 4 (1991) 187–197. [29] G. Holland, C.L. Bruyère, Recent intense hurricane response to global climate change, Clim. Dyn. 42 (3–4) (2014) 617–627. [30] Huffington Post (n.d.), Hurricane Sandy Crime. Available via 〈http://www.huf fingtonpost.com/tag/hurricane-sandy-crime/〉 (cited 11.11.14). [31] L.M. Hunter, Migration and environmental hazards, Popul. Environ. 26 (4) (2005) 273–302. [32] T. Kaplan, Cuomo Seeking Home Buyouts in Flood Zones, The New York Times, 03 February 2013. [33] M.F. Karim, N. Mimura, Impacts of climate change and sea-level rise on cyclonic storm surge floods in Bangladesh, Glob. Environ. Change 18 (3) (2008) 490–500. [34] T.R. Karl, J.M. Melillo, T.C. Peterson, Global Climate Change Impacts in the United States, Cambridge University Press, Cambridge, 2009. [35] D. Kniveton, Climate change and migration: improving methodologies to estimate flows, IOM Migration Research Series, No. 33, International Organization for Migration, 2008. [36] T.R. Knutson, J.L. McBride, J. Chan, K. Emanuel, G. Holland, C. Landsea, I. Held, J. P. Kossin, A.K. Srivastava, M. Sugi, Tropical cyclones and climate change, Nat. Geosci. 3 (3) (2010) 157–163. [37] C. Kousky E. Michel-Kerjan Hurricane Sandy, storm surge, and the National Flood Insurance Program: A Primer on New York and New Jersey. 2012 Resources for the Future.Issue Brief 12-08. [38] C.E. Landry, O. Bin, P. Hindsley, J.C. Whitehead, K. Wilson, Going home: evacuation-migration decisions of Hurricane Katrina survivors, Working Paper 07-03, Department of Economics, Appalachian State University, 2007. [39] F. Laczko, C. Aghazarm, Migration, Environment and Climate Change: Assessing the Evidence, International Organization for Migration, Geneva, Switzerland, 2009. [40] S. Laska, Homeowner adaptation to flooding: an application of the general hazards coping theory, Environ. Behav. 22 (1990) 320–357. [41] M. Leighton, X. Shen, K. Warner (Eds.), Climate Change and Migration: Rethinking Policies for Adaptation and Disaster Risk Reduction, SOURCE 15, UNU-EHS, Bonn, Germany, 2011. [42] M.K. Lindell, S.N. Hwang, Households' perceived personal risk and responses in a multihazard environment, Risk Anal. 28 (2) (2008) 539–556. [43] S.F. Martin, Climate change and international migration. Background Papers of the Study Team on Climate Change and Migration, German Marshall Fund, 2010, pp. 1–9. [44] A.H. Maslow, A theory of human motivation, Psychol. Rev. 50 (1943) 370–396. [45] NCADAC, see: National Climate Assessment and Development Advisory Committee, Federal Advisory Committee Draft Climate Assessment Draft Report, Available via 〈http://ncadac.globalchange.gov/〉, 2013 (cited 17.11.14).

228

A. Bukvic et al. / International Journal of Disaster Risk Reduction 13 (2015) 215–228

[46] R.J. Nicholls, S. Hanson, C. Herweijer, N. Patmore, S. Hallegatte, J. CorfeeMorlot, J. Chateau, R. Muir-Wood, Ranking of the world's cities most exposed to coastal flooding today and in the future, Executive Summary, OECD Environment Working Paper, 2007a. [47] R.J. Nicholls, P.P. Wong, V. Burkett, J. Codignotto, J. Hay, R. McLean, S. Ragoonaden, C.D. Woodroffe, Coastal systems and low-lying areas, in: M. L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden, C.E. Hanson (Eds.), Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK, 2007, pp. 315–356. [48] R.J. Nicholls, A. Cazenave, Sea-level rise and its impact on coastal zones, Science 328 (5985) (2010) 1517–1520. [49] NOAA, see: National Oceanic and Atmospheric Agency, State of the Coast, Available via 〈http://stateofthecoast.noaa.gov/vulnerability/welcome.html〉, 2013 (cited 04.11.14). [50] F.H. Norris, T. Smith, K. Kaniasty, Revisiting the experience behavior hypothesis: the effects of Hurricane Hugo on hazard preparedness and other selfprotective acts, Basic Appl. Psychol. 21 (1999) 27–47. [51] NCA, see: National Climate Assessment, Coastal zone development and ecosystems, Available via 〈http://nca2014.globalchange.gov/report/regions/ coasts〉, 2014 (cited 2.03.15). [52] NRC, see: National Research Council, Abrupt Impacts of Climate Change: Anticipating Surprises, The National Academy Press, Washington, D.C, 2013. [53] A. Oliver-Smith, Sea level rise, local vulnerability and involuntary migration, in: E. Piguet, A. Pécoud, P. de Guchteneire (Eds.), Migration and Climate Change, Cambridge University Press, Cambridge, 2011. [54] J.K. Raid, F.H. Norris, R.B. Ruback, Predicting evacuation in two major disasters: risk perception, social influence and access to resources, J. Appl. Sociol.

Psychol. 29 (1999) 918–934. [55] A. Siders, Hurdles for Gov. Cuomo's New York Buyout Program, Climate Law Blog, Sabin Center for Climate Change Law. Available via: 〈http://blogs.law. columbia.edu/climatechange/2013/02/05/hurdles-for-gov-cuomos-new-yorkbuyout-program/〉, 2013 (cited 10.06.15). [56] B. Smit, J. Wandel, Adaptation, adaptive capacity and vulnerability, Glob. Environ. Change 16 (2006) 282–292. [57] S.K. Smith, C. McCarty, Demographic effects of natural disasters: a case study of Hurricane Andrew, Demography 33 (2) (1996) 265–275. [58] K. Tierney, C. Bevc, E. Kuligowski, Metaphors matter: disaster myths, media frames, and their consequences in Hurricane Katrina, Ann. AAPSS 604 (2006) 57–81. [59] K. Trenberth, Uncertainty in hurricanes and global warming, Science 308 (5729) (2005) 1753–1754. [60] P.J. Webster, G.J. Holland, J.A. Curry, H.R. Chang, Changes in tropical cyclone number, duration, and intensity in a warming environment, Science 309 (5742) (2005) 1844–1846. [61] White House, The Hurricane Sandy funding needs. In a letter from Executive Office of the Presidents, Office of Management and Budget, to members of the House of Representatives, 2012. [62] J. Yin, D. Yu, Z. Yin, J. Wang, S. Xu, Modelling the combined impacts of sealevel rise and land subsidence on storm tides induced flooding of the Huangpu River in Shanghai, China, Clim. Change 119 (3–4) (2013) 919–932. [63] G. Yohe, J. Neumann, P. Marshall, Ameden, The economic cost of greenhouse induced sea level rise in the United States, Clim. Change 32 (4) (1996) 387–410.