The practical use of social vulnerability indicators in disaster management

International Journal of Disaster Risk Reduction 63 (2021) 102464 Contents lists available at ScienceDirect International Journal of Disaster Risk Reduction journal homepage: www.elsevier.com/locate/ijdrr The practical use of social vulnerability indicators in disaster management Erik Wood *, Monica Sanders, Tim Frazier, Phd Georgetown University School of Continuing Studies, Emergency and Disaster Management Master’s Program, 640 Massachusetts Ave NW, Washington, DC, 20001, USA A R T I C L E I N F O A B S T R A C T Keywords: Poverty Vulnerability Indicator model FEMA SoVI SVI Sendai As climate change focuses more frequent and intense disasters on vulnerable communities across the globe, mitigation and response resources need to be allocated more efficiently and equitably. Vulnerability assessments require time, skill, and cost money. In the United States (US), these assessments are mandated by the Federal Emergency Management Agency (FEMA) to qualify for federal funding. However, new trends in the literature clearly question their practical value. This study begins with a focused literature review which demonstrates that popular social vulnerability indices that create a single vulnerability score can diminish the significance of a lone variable, overlook the relevancy of all interconnected variables, and can result in contradictory policy recommendations. Next, existing case studies that used popular vulnerability assessment frameworks were compared to maps generated of the same study area that employed the single variable of poverty. These case study comparisons demonstrated how considering this greatest common variable among different vulnerable groups can often - quickly, efficiently, and inexpensively - reveal close to the same county and sub-county level community vulnerabilities detailed in costly assessments. Finally, a national survey of emergency managers was conducted to determine how much current social vulnerability indices were actually governing the ongoing distribution of resources to the communities being served. Results indicate that, while these indicator models may be underused nationally, those who do find them effective tend to be from higher income areas. This study questions the practical value of these indices for emergency management practice in the US and for meeting the goals of the Sendai Framework and other compacts. 1. Introduction Differential vulnerability, where different groups in varying spatial contexts are not equally predisposed to risk from a natural or humanmade disaster, is well established in the literature [1]. The term risk in this context involves the probability of a disastrous event harming “what humans value” as defined by their degree of vulnerability [2]. The International Federation of the Red Cross (IFRC) defines vulnerability as a state most often associated with poverty and based on the relative ability of an individual or community to cope with, resist, and recover from a hazard [3]. The term social vulnerability connects the characteristics of that individual or group to their inability to resist the forces of the hazard [76]. Since the social vulnerability index (SoVI) was introduced [4], there have been several similar vulnerability frameworks developed globally [5] which employ a composite or aggregation of multiple vulnerability indicators to derive a single measure of vulnerability. Some of those include the Spatially Explicit Resilience-Vulnerability (SERV) model [6], the Natural Hazards Index developed at Columbia University [7], FEMA’s National Risk Index [8], and the Health Vulnerability Index [9]. After SoVI, one of the next most widely used indicator models is the Center for Disease Control’s (CDCs) social vulnerability index (SVI) [10]. The main focus of this project was limited to these two widely used and impactful indices. The use of SoVI and SVI is facilitated by FEMA grant programs. This includes the Emergency Management Performance Grants (EMPG), the Homeland Security Grant Program (HSGP), and the Urban Areas Security Initiative (UASI), which require these kinds of vulnerability assessments as part of their funding application. Additionally, much of the post-disaster assistance outlined by the Stafford Act of 1988 [11] available to communities in the US is contingent upon having a current Hazard Mitigation Plan (HMP) which includes an assessment of vulnerabilities of the population being served [12]. While there is little peer-reviewed research on what county and subcounty resource allocation is directly driven by social vulnerability indices, one can find related academic studies and state or county level technical documents. Flanagan et al. [13] is an example of a typical * Corresponding author. E-mail addresses: [email protected] (E. Wood), [email protected] (M. Sanders), [email protected] (T. Frazier). https://doi.org/10.1016/j.ijdrr.2021.102464 Received 17 May 2021; Received in revised form 9 July 2021; Accepted 10 July 2021 Available online 17 July 2021 2212-4209/© 2021 Elsevier Ltd. All rights reserved. E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 academic study in this space. Cited over 700 times, this well documented research examines a historic disaster (in this case Hurricane Katrina) and, using SoVI, determines what could have been done differently and how vulnerability could be mitigated in the future in the face of similar shocks. Directed by the 2013 FEMA Mitigation Planning Toolkit, many states, like Georgia, also use SoVI to develop a snapshot of the social vulnerability conditions across their counties and documented in their HMPs [14]. “This comparison can be used by state, regional, federal authorities in the allocation of pre and post disaster resources to better ensure that these are delivered where most needed” [14]. Many states and counties in the US opt to hire private consultants to meet these federally mandated assessment deadlines. Meanwhile, the efficacy and efficiency of SoVI and other popular social vulnerability assessments, like SVI, have been challenged in recent literature [5, 15–22]. Yet the number of studies that challenge the validity of leading vulnerability assessments is still extremely low [19]. “SoVI and the SVI are the most prominent social vulnerability configurations and are being promoted for use by public health officials and planners to identify socially vulnerable places and populations. Our empirical analysis raises questions about their construct validity” [19]. Within this context, this study will explore how much these vulnerability indices are facilitating the actual distribution of disaster mitigation and recovery resources to the most vulnerable populations. distribution of disaster mitigation and response resources at the county and sub-county level? Search terms included: vulnerability, disaster, index, and/or SVI and/or SoVi in multiple Booleen arrangements. Articles considered “slightly relevant” or “likely relevant” were read completely for further assessment. Initially, 19 abstracts and conclusions of potentially related articles were reviewed. Next, only 11 core, peer-reviewed articles were retained for being significantly relevant to this subject matter. This confirms the scarcity of research on this topic mentioned in Rufat et al. [19]. While several of the 11 core citations also discuss the key variable of poverty (e.g. Refs. [16, 27], 13 additional citations were added to further establish the connection between vulnerability and poverty. Finally, additional citations were considered and added by searching those core articles’ citations to look backward in time for additional relevant sources and by looking more recently at where those core works had themselves been cited. Three citations were added outside of this matrix to support the use of a focused literature review for this type of study. With two expert book author additions and two contextual citations (e.g. Ref. [4] this resulted in a total of 30 citations related to the focused literature review. 2.2. Case studies methodology The case studies in section 4.0 were chosen based on being readily available and that they employed one of the more popular social vulnerability indices previously discussed. In each case, areas of these studies were enlarged to accommodate comparison at the graphic size generally suited for publishing in an academic journal. Through visual inspection of each map product, the goal was to prove or disprove whether the single variable of poverty would produce similar assessment results that the multi-faceted and multiple variable SoVI and SVI-style studies produced. While the visual comparison approach to model convergence has been criticized as a means of in-depth analysis, this type of analysis has been shown to support basic correlations (e.g. Refs. [28–30], that then require support from other qualitative and/or quantitative sources to increase plausibility. Therefore, this method was deemed appropriate for the purposes of exploring a potential correlation between the Poverty Vulnerability Map (PVM) outputs and widely known outputs. Additionally, the single variable PVMs had to be produced in 15 min or less using ESRI’s ArcGIS Online (AGOL) by a modestly skilled operator and not by an advanced technician using ArcGIS Pro or ArcMap. Authors reasoned that this more closely resembled the average skill, technology, time availability, and budget constraints experienced by existing emergency management bodies at the county and sub-county level in the planning and mitigation phases. A total of six case studies were conducted. In each of the PVM case studies, two complimenting data layers were used. This included “Low Income Community Census Tracts - 2016 ACS” which contains US Census data from 2012 to 2016 at the Census Tract level related to poverty and median family income. The second, complimentary data layer was “ACS Poverty Status Variables Boundaries” which contained 5-year poverty estimates (2014–2018) by tract, county, and state. Here again, these layers were chosen, not with experience of an advanced technician, but by searching data layers within AGOL’s Living Atlas content and using the search term “US Census, Poverty.” These two layers were the first two results in that search and their metadata indicated good quality. While timeframes for the poverty data were close to those in the case studies, our goal was to use data that was the same vintage or slightly newer as poverty trends tend to continue in the same locations [31]. In each case study, the author/operator attempted to quickly adjust the color range of the PVM to, as closely as possible, resemble the existing social vulnerability map. This would accommodate comparison without sacrificing the 15 min or less rule. Finally, it is important to note that each case study was chosen for its SoVI or SVI-based map result of an easily defined area in the US to determine if the 15-min PVM would display vulnerability patterns that were highly similar, somewhat 2. Methodology 2.1. Focused literature review methodology This research is based on a mixed methods approach synthesizing data outputs from both qualitative and quantitative data to answer a research question [23]. This included a focused literature review, a comparison of single indicator versus more complex vulnerability mapping outputs, and a nationwide pilot survey targeting practicing emergency managers (see Fig. 1). The introduction to this study includes a focused, qualitative literature review of relevant work. This method is common within peerreviewed literature to establish the plausibility of new ideas, approaches, or to encourage additional research (e.g., Refs. [24–26]. The main research question for this study was: What is the accuracy of existing vulnerability assessments and how are these studies influencing the actual Fig. 1. Data sources and synthesis. 2 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 similar, or not similar at all when compared to the much more complex frameworks. The intent is not to be critical of any of these existing studies or their contributions to the literature. In fact, one of the six studies [19] is also a supporting citation for this study. Furthermore, the purpose of this study is not to argue against the use of the multi-indicator approach. However, analyzing the speed, access, and low-cost when employing a dominant, single variable like poverty can and should make the argument for improved utility of and access to vulnerability assessments - especially within the most vulnerable and poorest communities. Table 1 The four core survey questions and Likert-type, five option response range. 2.3. Survey methodology A national email survey directed at public agency emergency managers working at the state, county, or sub-county level was conducted. Amassing email contacts where the individual either carried the title of emergency manager (EM), emergency services coordinator or similar was the goal. This included adjoining titles such as director, deputy director, chief, and specialist. Filtering potential recipients to make sure the person is an emergency manager or acting in this capacity for that area was a priority. For example, to obtain EM email addresses from California, a Google search led to the CalOES Regional contacts page [32] where 41 current email addresses were collected. Two potential email addresses from this website were not added to the list as they carried the sole title of Government Program Analyst which did not qualify as likely to carry EM decision-making responsibilities. A total of 2128 contacts were gathered from every state across the United States using this method. Next, the publicly accessible directory of members in the International Association of Emergency Managers (IAEM) website was filtered using the aforementioned criteria and 1074 current email addresses were added to the list. These were compared to the other list of Google search contacts using Qualtrics software to ensure duplicate emails would not be sent out which is likely to lower participation. The final contact list, without duplicates, totaled 3108 contacts. This was a cross sectional survey as all data were collected at once using survey software from Qualtrics to distribute the survey via email to the 3108 total contacts. To avoid compromising the data with social desirability (participants wanting to be perceived well), all questions were answered anonymously. This anonymous method directed at prescreened professionals was also considered low risk without ethical challenges opposite the institutional review board (IRB) standards [33]. Two gateway questions began the survey to further qualify respondents. The first of these questions asked the participant to simply confirm their role in emergency management. The second gateway question (“Does your agency use a community vulnerability assessment tool like SoVI or SVI?“) employed the Qualtrics skip logic feature and only allowed participants who answered in the affirmative to participate in the next four core survey questions. Those who answered “No” or “Do Not Know” had completed their survey. This gateway question helped ensure that those answering the next four key vulnerability assessment questions were not, for example, responding in the negative answer range because they did not use this assessment or did not know what it was. The Likert-based structure to the survey questions is common when measuring attitudes toward a specific subject [34]. Five response options were provided so that respondents had a neutral response choice and would not feel they were being compelled to either answer positively or negatively [35]. All four Likert-type questions revolved around a single theme of “vulnerability assessments” and avoided extreme language that could elicit a tendency toward a central or neutral response (see Table 1). The intent was to make this as short a survey as possible (rated at 1-min) to increase response rates from a historically and presently extremely busy group of professionals. The simplicity of the survey was intended to yield easily interpreted results that would help answer the main study question stated in section 2.1. Qualtrics did anonymously provide Internet service provider (ISP) address respondent location data (latitude and longitude) which would facilitate further Four core Likert-type survey questions Five response option range How likely are you to refer to your current vulnerability assessment on a monthly basis? How likely are you to refer to your current vulnerability assessment to help allocate resources to the community you serve? How likely would you support your agency budget paying for your current vulnerability assessment if it was not required by DHS/FEMA? If you do use a vulnerability assessment (like SoVI or SVI), please rate how effectively it helps you allocate resources to the community you serve from “very effective” to “not effective at all" Very unlikely to Very likely Very unlikely to Very likely Very unlikely to Very likely Not effective at all to Very effective analysis of the responses. The assumption was made that, especially during the height of the coronavirus travel restrictions when the survey took place, US emergency managers ISP location would likely mimic the location of the community they serve in. 3. Focused literature review The current state of vulnerability assessments accepts the aggregation of indicators without knowing which variable is most associated with that community’s vulnerability or which variable may actually be masking this correlation, potentially skewing the equitable distribution of resources which is defined here by three main categories including financial, policy, and technical assistance [36]. Subjectively weighting one of the 30 or more socioeconomic variables like age, residential property, or occupation [37] higher than another, potentially more locally relevant variable, like poverty, can produce misrepresentative results [5,19]. Additionally, it is not evident which variable contributes most to community vulnerability or that one vulnerability assessment is compatible with another which could be a significant impediment to the Sustainable Development Goals of the Sendai Framework [5]. This same study found conflicting outputs at a national and international level between comparative vulnerability index studies (e.g., SoVI or SVI) that use qualitative methods and self-assessment studies that use qualitative inputs. While composite indicators used in vulnerability indices may provide a smaller scale digest of complex issues, they can also result in misleading or overly simplistic policy conclusions [15]. The amount of data needed to compile vulnerability indices may be out of reach of many local resilience bodies without the expense of a consultant or team of consultants. Supporting mapping software expenses (direct or via consultant fees), from an industry leader like ESRI, can exceed US$12, 000 with annual maintenance fees of US$3000 or more [38]. However, statisticians may consider an aggregation of indicators that result in a single measure to carry minimal practical significance. Saisana et al. [15] describe the ease with which a single output like this to quickly form policy is irresistible. Others question the applicability of these indices for hazard mitigation. “These benefits of indices make it likely they will continue to be attractive to stakeholders, practitioners, and decision makers, despite counterarguments that generalized aggregated indices are too blunt an instrument to be used for specific hazard mitigation interventions” [16]. Chang et al. [1] determined that the assessment of vulnerability factors has shown little value for local disaster managers attempting to equitably allocate resources. Ilbeigi and Jagupilla [21] concluded that SoVI has “critical shortcomings” and “might be misleading.” These authors also note that not all vulnerable populations occupy physically vulnerable locations. Large, scenic US coastal cities, for example, tend to attract higher income groups despite the physical exposure to shocks [21]. This relates to the concept of resilience which is defined here as ‘‘the ability to prepare and plan for, absorb, recover from, and more successfully adapt to adverse 3 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 events” [39]. This also speaks to two divergent frameworks of disaster resource distribution: equal versus equitable. Higher-income individuals occupying scenic coastal areas have access to insurance, bureaucratic and political agency, and attractive options for temporary or even permanent living alternatives (e.g., second homes) [31]. Therefore, equitable emergency management resource allocation may need to give more to those who appear to be impacted less than those who absorb the greater physical impact but are intrinsically more resilient due to their income. Equal distribution of resources (or even weighted to address the ‘greatest physical impact’), on the other hand, may disproportionately benefit those already more resilient while areas of ‘lesser physical impact’ may represent the greatest assault on short and long-term, whole-community resilience. Today, however, resources are often directed towards the greatest physical damage and not necessarily the most vulnerable [6,40] which implies vulnerability mapping may not be connecting to disaster management efforts on the ground or that the mapping is flawed. “While we found flood recovery assistance to be strongly associated with physical damage overall the relationship was more tenuous in places with higher social vulnerability” [40]. The current state of social vulnerability literature has numerous studies that demonstrate the inequitable distribution of resources through all phases of the disaster management cycle [40]. Validating the efficacy of social vulnerability indices has also proven to be difficult, yielding mixed or inconclusive results [16]. Subjectivity is also inevitable since all models require the model builder to choose between variable levels and weighting [16]. Additionally, different assessments of the same location have been shown to produce noticeably different results [22]. “To a naïve observer, these latent variable methods might seem like voodoo, that is, via the magic of statistics, one can generate a quantitative description of something that they otherwise could not observe” [22]. A more complex index can, in fact, be counterproductive to the efficient and equitable distribution of mitigation resources when a critical variable, like poverty, is not singled out for analysis. This variable might relate more clearly to the type of vulnerability an organization is attempting to address [17]. “SoVI® implicitly weighs old age, making some tracts more socially vulnerable than they should be. CDC SVI exhibits homogenous patterns of social vulnerability, so uniquely vulnerable tracts may go unnoticed” [17]. Higher income directly correlates to increases in life expectancy [41] which must also be considered when weighting the age variable. The coronavirus pandemic has demonstrated the interconnectivity of variables and how that can impact index outputs. Racial bias, for example, is associated with age-based generalizations related to risk. In the US, the age-based pandemic mortality rate for Blacks was 3.4 times higher than for Whites [42]. Pandemic data are also demonstrating a strong correlation between age, race, and income - revealing distinct differences in infection rates and deaths at the county and zip code level [43]. “Use of distinct social metrics (rather than summary indices) is important to guide action: data on household crowding, poverty, and residential segregation by race/ethnicity and income are all informative and needed to communicate the social burdens of COVID-19” [43]. And while SoVI and SVI consider income as one of 20, 30, or even 40 other variables including age, race, and income, they do not, for example, consider income distribution within neighborhoods and households. A single parent household will have a different level of vulnerability than a dual earner household. A single-income household led by a Latina or African American woman will be more vulnerable than most other households located in the same area [44]. Furthermore, neither popular index (SoVI or SVI) includes homeless populations which can greatly impact response and recovery [17]. Frazier et al. [20] points out two key SoVI drawbacks. First, exposure, sensitivity, and adaptive capacity are not included and second, SoVI is not differentially weighted for all indicators that, in turn, impact the resulting vulnerability score [20]. And yet all of these types of indicator models, including others like the SERV framework, are themselves vulnerable to the subjective weighting of variables which can result in inaccurate geospatial assessments of community vulnerability. These models also require a complete buy-in from the emergency management team that the aggregation of vulnerability factors produces a metric that speaks to the specific block-level risk and resilience goals of that community over, for example, a single variable – like poverty. Barrett and Constas [74] maintain that the ability of an individual, household, or community to avoid poverty is the single greatest component of resilience. “Our work complements a flurry of recent notes, white papers, and policy briefs on resilience, all offering measurement tools, policy and programming prescriptions, etc., but conspicuously without any explicit theory of resilience as might be applied to the lives of the poor” [74]. This connection between poverty and vulnerability opposite the threat of disasters has already been well established in the literature (e.g., Refs. [3,45–48]: [27,31,49–53]. “Poverty is an important aspect of increased social vulnerability because of its direct association with access to resources which affects coping with the impacts of disasters” [27]. The ongoing coronavirus pandemic is providing further, visceral evidence about the causal relationship between poverty and communal vulnerability. In one national pandemic epicenter, Los Angeles County Public Health Department reported mortality rates in areas of 30%–100% poverty that are 3.7 times higher than areas of less than 10% poverty [54]. Glynn and Fox [55] found a direct causal relationship between rent increases and increases in homeless populations in several major US cities. “Communities where people spend more than 32% of their income on rent can expect a more rapid increase in homelessness” [56]. Today there is no place in the United States where full-time, minimum wage earners can afford a two-bedroom rental. Ninety-five percent of this same group cannot afford a one-bedroom rental [57]. In terms of disaster mitigation and response, an increasing homeless population directly translates into an increase in whole-community vulnerability [56]. A Rand Corporation study revealed that, since 1975, $2.5 trillion per year (in 2018 dollars), which is approximately 12% of GDP per year, has been redistributed from the bottom 90% of earners to the top 1% of Americans [75]. Between 1944 and 1974 all worker incomes in the US rose at an even pace (or 100%) with US GDP. The rising economic tide essentially lifted all boats. Since 1974, however, those same incomes only increased by 17.4% of GDP [75]. A review of related literature by Hallegatte et al. [51] found that disasters consistently increase poverty levels and reduce income levels, and that the field of disaster management should be a part of poverty reduction policy. For example, Hallegatte et al. [58] states that “in relative terms, poor people always lose more than non poor people from floods and storms” [58]. The processes that exploit poverty during disasters will only be exacerbated by climate change [48,58,59]. However, the inability to connect post-disaster economic impacts, including increased poverty and homelessness, to disaster management mitigation practices prevents any kind of meaningful resiliency evaluation of those same practices [60]. 3.1. Evaluation of law and policy frameworks The complexity of matching an accurate measure of vulnerability with the appropriate tools to achieve successful mitigation is complicated by the law governing hazard mitigation funding and operations. For example, Public Law 106–390, or the Disaster Mitigation Act of 2000, brings consideration of social vulnerability into hazard mitigation planning. This bill amended the Stafford Act with the intent to streamline predisaster mitigation and control and/or lower the costs of federal assistance. In drafting the statute, policy writers placed “effective community partnerships’’ and reduction of loss of life, property, economic impact and disaster assistance costs as a priority. The writers also heavily weighted the term “small impoverished communities’’ defined in the law as “… a community of 3000 or fewer individuals that is economically disadvantaged, as determined by the State in which the 4 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 community is located and based on criteria established by the President” [61]. Reading these sections together, along with the Senate Report on the legislation, it is clear that FEMA has funds designated to use technical tools to improve the assessment of community vulnerability, generally speaking and specifically for these kinds of communities (Senate Report PL 106–390). Based on the interpretation of the Disaster Mitigation Act of 2000 and other barriers noted herein (e.g., time, skill, and cost), where social vulnerability assessment tools are needed, the agency should elect to use the most specific, time sensitive, accessible, and cost-effective tool available. This is particularly critical when working with those which qualify as “small impoverished communities.” Six years later, Congress would deepen its focus on impoverished communities and the broader issue of social vulnerability. In the aftermath of Katrina, the “Post-Katrina Emergency Reform Act of 2006” [62] was written to address the harsh lessons learned about poverty, vulnerability, and governance from that event. In particular, the legislation focused on understanding and serving the disabled, providing temporary “social safety net” assistance related to housing, rental payments, and unemployment, as well as eliminating the denial of other government benefits based on receiving this assistance. The bill added additional emphasis on assistance to communities, both financial and technical, while also ensuring more agency involvement in recovery. In short, addressing poverty and vulnerability became more important in emergency management [62]. When considering the need to assess communities’ that have been hit hard by a disaster, assessors have to consider pre-existing vulnerabilities alongside other damage caused by the event. Expediency, availability, and cost-effectiveness should dominate the consideration of assessment tools. Knowing this is the intent of the law, how does the use of these relatively expensive tools realistically fit within the framework of disaster mitigation? This is a question not only being posed at the state and local level in the United States, but at the international level as well. Referring to the growing sets of social vulnerability assessment tools, the United Nations Development Programme (UNDP) offers a cautious description of their implementation in its own handbook: analytical tools and in planning generally. This is an area ripe for further research as this matrix tends to support that existing poverty and vulnerability begets sustained or even increased poverty and vulnerability. In the international context, the term “impoverished” could be about: funding, income levels, the harms of poor governance regimes in the country, as well as neo-colonial attitudes within aid agencies. The outcome for those managing and planning for catastrophic events will be similar in terms of having to make difficult decisions about tools and planning. The specific contextual analysis, again, is ripe for further research. Here, we discuss whether, with respect to SoVI and SVI, renewed consideration about the variables, availability, and uses will yield uses and outcomes closer to the Sendai Framework goal of “freely available and (widely) accessible”. 3.2. SoVI and SVI The home of the Social Vulnerability Index (SoVI) is the Hazards and Vulnerability Institute at the University of South Carolina. This institute also houses the International Centre of Excellence in Vulnerability and Resilience Metrics (ICoE-VaRM). The website indicated that it “collaborates with the International Digital Belt and Road (DBAR) Research Program through their Center of Big Earth Data for Coasts (ICoE-Coasts), NOAA’s Carolinas Integrated Sciences and Assessments (CISA) program, and the University of South Carolina’s Artificial Intelligence Institute.” Just within the University of South Carolina, there is an abundance of opportunity to proliferate the index for academic purposes. In theoretical discussions, the indicators are used to broadly describe social vulnerability in the context of resource mapping or to help with machine learning and artificial intelligence development. SoVI is an index created by Dr. Susan Cutter and other academics at the University of South Carolina’s Hazards and Vulnerability Institute. It was developed as “Social Vulnerability to Environmental Hazards” and intended for use by stakeholders at the county level [17]. This index is about managing disaster risk or developing tools for disaster risk reduction via the lens of social vulnerability. The CDC’s SVI was created by the Centers for Disease Control Agency for Toxic Substances and Disease registry to identify populations that would need more resources in the event of a disaster or emergency (CDC’s Social Vulnerability Index (SVI), 2020). This is a public good, released freely online and updated in a biannual report. It uses 15 variables from the U.S. Census. Those variables are ranked by impact on vulnerability to come up with a vulnerability index and subsequent ‘scores’ for communities. Most of the data used by these two vulnerability assessment frameworks comes from the US Census, which is common but not perfect, demonstrating margins of error consistently over 10% (US Census Bureau, n.d.) and often over 30% [22]. The key differences are that SVI is a public good and SoVI comes with a cost; one is about resource allocation (or other creative uses by those using the information) and the other is about disaster risk reduction. Despite their broad use, both indices have inconsistencies as demonstrated in recent literature and detailed in section 3.0 of this study. “A number of issues and challenges arise in defining and mapping the socially vulnerable population within climate change adaptation and disaster risk assessment frameworks. The variety of parameters used to determine social vulnerability includes, but is not limited to, income disparity, gender, age, disability, language, literacy or family status. However, personal characteristics can be linked to vulnerability, but not define it. Group approach is widespread because it is easy to administer and use for targeting the population, but it largely ignores the internal heterogeneity of groups” [63]. Further, the Sendai Framework for Disaster Risk Reduction [64], which encompasses both the risk reduction and vulnerability planning goals of both SoVI and the CDC’s SVI, requires actors “… to systematically evaluate, record, share and publicly account for disaster losses and understand the economic, social, health, education, environmental and cultural heritage impacts, as appropriate, in the context of event-specific hazard-exposure and vulnerability information.” Additionally, this framework requires stakeholders to “make non-sensitive hazard-exposure, vulnerability, risk, disaster and loss-disaggregated information freely available and accessible, as appropriate” [64]. Looking at the international perspective offered from both organizational and principled perspectives, it is important to ask about the heterogeneity of the variables as suggested by the UNDP, but also to question the meaning of “freely available and accessible.” For those in emergency management related positions in the United States, particularly within these “small impoverished communities” contemplated in the Disaster Mitigation Act of 2000, this could be a budget question. Current emergency management and disaster science thinking rarely delves into the commonsense nexus between poor communities, lack of tax revenue, and the difficult decisions the two create for emergency managers when choosing 4. Results: case study comparisons 4.1. Case study 1: Roane County, Tennessee In their article titled Developing a Climate-Induced Social Vulnerability Index for Urban Areas: A Case Study of East Tennessee, Omitaomu and Carvalhaes [65]. developed a social vulnerability assessment using the CDC’s SVI model. The authors ultimately use this assessment to analyze climate change vulnerability in the same area. See Appendix 10.1 for more detail on this study’s methodology. In terms of resource allocation, Omitaomu and Carvalhaes [65] focused on the equitable development of greenspace within more vulnerable communities. The study also points out that green space development can have the opposite effect on 5 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 poorer communities by inviting gentrification and higher taxes which ultimately displaces those communities [65]. Fig. 2 shows an excerpt from one of the counties this study analyzed during their SVI map generation. Scores closer to 1 represent greater social vulnerability while those closer to 0 represent lower vulnerability. With reference to Fig. 3, results of this case study indicate an unequivocally strong similarity between, not only the two distinct areas of high vulnerability within Roane County, Tennessee but also those areas of lesser vulnerability. There were no notable areas of dissimilarity. case, the 15-min PVM displayed patterns of vulnerability that were highly similar to the more complex models. In some cases, small areas on the PVM only showed patterns that were somewhat similar, but this was generally attributable to the more refined gradations that the SoVI/SVI map legends showed. When, for example, the top two highest levels of vulnerability were matched to the PVM’s Low Income category, any minor dissimilarities would lessen. Notably, there was not a single PVM created that displayed areas that were not similar at all to the more complex map products. This demonstrates the significance of the poverty variable when determining vulnerability while raising concerns (noted in the literature review) that aggregated indices may average out this significance. This is combined with the concern that the related policy frameworks (section 3.1) are encouraging the use of these indices. 4.2. Case study 2: US regional SoVI In their article titled Risks of sea level rise to disadvantaged communities in the United States, Martinich et al. [66] used SoVI in their study of four key US regions to analyze social vulnerability opposite the effect of climate change and sea-level rise (SLR). Fig. 4 shows one of four regions analyzed using SoVI. See Appendix 10.2 for more detail on this study’s methodology. In terms of resource allocation, Martinich et al. [66] attempt to determine if populations displaying higher social vulnerability have the ability to harden their communities against SLR or if relocation may be a more viable use of available resources. “In effect, areas at risk of SLR, but with low levels of social vulnerability, should be able to effectively respond to this risk by fortifying shores or nourishing beaches, while more socially vulnerable populations would be more likely to have fewer resources within their communities to respond in this manner [66]. Fig. 5 shows the same region using the 15-min PVM method. Results of this case study again indicate a strong similarity between patterns of vulnerability seen within this regional scale map product and no obvious areas showing a strong dissimilarity. In particular, areas south of the Philadelphia, Baltimore, and Washington labels are strikingly similar. The SoVI map product represents a finer granularity in vulnerability classifications, showing a five-level vulnerability range whereas the related PVM only shows two (Low Income and Not Low Income). However, the similarity tended to increase when the highest two SoVI vulnerability levels were compared to the Low Income areas from the PVM. A total of six of these case studies were conducted. The two comparisons chosen for publication had the highest resolution and represented the results of all six case studies. The other four included: Flanagan et al. [67,68]; Rufat et al. [19]; and Lehnert et al. [69]. In each 5. Results: survey OF US emergency managers A nationwide survey was emailed to 3108 emergency managers or those acting in an emergency management role for their US state, county, or sub-county public agency. From this contact list, 267 contacts “bounced” indicating those emails were no longer valid or active. This left a total survey pool of 2841 contacts from which 485 responded equaling a 17% response rate. The first two gateway questions further reduced the participant size. Question 1 (Can you confirm that you are an emergency manager, coordinator, or work in a related role for the community you serve?) yielded 19 “No” responses which left 466 to potentially respond to the four core survey questions. However, 100 potential respondents dropped out of the survey after Question 1, possibly realizing they did not qualify or for other reasons like connectivity. This left a net pool of 366 participants. Question 2 (Does your agency use a community vulnerability assessment tool like SoVI or SVI?) yielded 247 “No” responses and 52 “Do Not Know” which, combined, represented 81% of the remaining participants. This left 67 US EM professionals and seven of those failed to continue, leaving a group of 60 for the final four questions. This moved this survey classification to that of a pilot survey which has been shown to have value in advancing the literature (e.g., Refs. [70,71]. Location data provided by Qualtrics were overlaid onto the response locations from the questions in Table 2 to attempt to identify any poverty-related trends. Again, using AGOL software with the Low Income Community Census Tracts - 2016 ACS data layer (reference section 2.2), each response location was analyzed to determine if responses came Fig. 2. Roane County, Tennessee vulnerability using SVI from Omitaomu and Carvalhaes [65]. 6 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 Figs. 3. 15-min PVM of Roane County, Tennessee that matches Fig. 2 study area. Fig. 4. One of regional SoVI maps produced by Martinich et al. [66]. 7 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 Figs. 5. 15-min PVM of US region that matches Fig. 4 study area. from a “Low Income” area or a “Not Low Income” area. Trends were recognized in Questions 4 through 6 (See Table 3). While the sample size of this pilot survey is not robust enough to support standalone claims, it is reasonable to infer from the trends noted in Table 3 that the popularity of the available vulnerability assessments tends to lean toward areas that are “Not Low Income.” This result is supported by the literature review that suggests that the nationallydriven vulnerability framework may not be supporting the most vulnerable communities. These survey findings also support the call for a more comprehensive nationwide survey. practice of emergency management within poorer communities. Meanwhile the efficacy of the most popular indicator models has also come into question. “Present practice has two key limitations that may restrict their use or potentially lead to poor decisions being made in their implementation – low use of direct measures of disaster resilience and low use of sensitivity and uncertainty analysis” [5]. The aggregation of multiple variables can dull the complicity of a single one. All indicator models are susceptible to the subjectivity of input and weighting of variables [22] which can result in conflicting mitigation policy recommendations based on the framework that was used [18]. As stated earlier, the aim of this study is not to discount the historic and potential future value of the indicator approach. Before the advent of these models, social vulnerability was rarely discussed and, moving forward, their value could be more widely utilized with their limitations in plain sight. However, it is clear from the literature alone that there needs to be better utility of, and access to, vulnerability assessments - especially within the most vulnerable communities. The strong correlation between poverty and social vulnerability was not only evident in the literature but also in the case study map comparisons from section 4.0. The PVM examples executed in a short time span by a moderately experienced operator demonstrated how significant a single variable can be in determining vulnerability. The operator experience and time allotment also attempted to address (or at least acknowledge) the access poorer communities have to mitigation resources like more complex, time consuming, and costly vulnerability modeling. Simply considering poverty as a common denominator to many forms of vulnerability in many communities may ironically lead to a more freely available and widely accessible means for those very communities to mitigate their own vulnerability. The PVM case studies were not conducted to advocate for whole community risk reduction based on the single variable of poverty but to provide a lens from which 6. Discussion This study conducted a focused literature review, reviewed legal frameworks, produced comparative geospatial vulnerability map case studies, and conducted a nationwide pilot survey of emergency management professionals. The results of the focused review clearly show flaws in the current vulnerability assessment indices that are federally mandated. Whether it is the issue of variable heterogeneity, oversimplified outputs discussed by the UNDP, or the lack of free availability and wide access called for by the Sendai Framework, the literature points to problems with each. Next, previous research supports a troubling, and increasingly plausible, connection between the lack of tax revenue in poorer communities leading to less access to tools by emergency managers to assess and mitigate vulnerability. FEMA may be unknowingly contributing to a cycle that moves the true potential of the indicator approach further and further away from influencing mitigation in the poorest, most vulnerable communities. The maxim that vulnerability begets vulnerability may, in turn, be supported by FEMA grant programs that encourage the inclusion of vulnerability assessments (often by paid consultants) that are less and less accessible to the 8 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 might be underused nationally. Additionally, those that were using the SoVI or SVI models and found it to be either effective or very effective tended to be from areas that were “Not Low Income.” In fact, Questions 4 through 6 demonstrated this trend where communities that were “Not Low Income” tended to be the ones favoring these popular assessment tools. (see Table 3). While the literature supports the plausibility of this connection, additional research that captures a larger sample size from the emergency management community is required to verify and add depth to this relationship. As the first-known survey of its kind, the overall results also point to the need for a more robust sample size that might include follow up, semi-structured interviews to determine, for example, in what manner vulnerability assessments are being used and how that use benefits the vulnerable populations in that community. This could perhaps be conducted anonymously by FEMA but as part of the grant program process discussed in section 3.0. Assessments should drive mitigation and the simplicity of access to the poverty variable, as shown in this study, points to how key variables can be accessed and used for mitigation in the most underserved and poorest areas in the US. While looking at a single variable may be less complex, time consuming, and costly, it also ensures that this variable is not averaged out. SoVI and SVI are not necessarily the wrong tools but their normalization within the DHS/FEMA agency structure may also be normalizing higher income community access to less accessible indicator driven assessments. This inadvertently leads to vulnerability in poorer communities begetting higher vulnerability to future shocks in those same communities. Table 2 Aggregated core question results. Question Number Question Text Very Unlikely + Unlikely Neither (Neutral) Likely + Very Likely 3 How likely are you to refer to your current vulnerability assessment on a monthly basis? How likely are you to refer to your current vulnerability assessment to help allocate resources to the community you serve? How likely would you support your agency budget paying for your current vulnerability assessment if it was not required by DHS/ FEMA? Question Text 38% 29% 33% 18% 15% 67% 35% 12% 53% Not effective at all þ Not very effective Neither (Neutral) Effective þ Very Effective If you do use a vulnerability assessment (like SoVI or SVI), please rate how effectively it helps you allocate resources to the community you serve from “very effective” to “not effective at all" 13% 28% 59% 4 5 Question Number 6 7. Conclusion Thinking through the variety of uses for which these indices are deployed and why, in addition to suggesting more accessible routes to their benefits is one approach. Another is to review the policies which incentivize the use of these indices and suggest changes that can help socialize and operationalize some alternative approaches. Many times, they are used in mitigation planning and for participation in FEMA’s Hazard Mitigation Grant Program, which in 2020 was renamed the Building Resilient Infrastructure and Communities (BRIC) Grant Program. The program funds mitigation projects as well as capacity building and management costs. In the mitigation eligibility section, identifying and developing programs that impact “underserved” communities and populations rates high with the first updated activity language reading, “Updating the risk and vulnerability assessment based on new information, including supporting studies, such as economic analyses, mapping, risk assessment, and planning” [72]. Much like the preceding program, this one will require or incentivize the use of vulnerability assessment tools for the grant program itself and to be incorporated into other planning activities. Thus SoVI, CDC’s SVI, and similar tools are in play in this program as a matter of policy. FEMA could add flexibility into the program by moving beyond a focus on risk and vulnerability to one which considers equity, as it has done in the “Handbook on Expanding Equity” which focuses on operationalized equity internal to the agency and externally. This handbook acknowledges the CDC’s SVI as part of their approach [73]. Looking at this framework, there are suggestions which address some of the root causes of the vulnerabilities and how to overcome them. One suggestion is simply working in concert with community leaders. The agency should encourage this engagement, which will reveal more practical truths about vulnerability than mapping and point data alone, as a qualifier for the grant program. For applicants in disaster prone states where assessments have consistently been done in the past, and are still within a reasonable period of the last assessment (10 years for census based indicators); those applicants should be allowed to base assessments on factors that reflect the greatest community need and be able to qualify. For example, a community in a hurricane prone state that was also hit hard by COVID-19 may want to view hazard mitigation from the lenses of increased poverty because of the pandemic’s Table 3 Low income versus not low income response trends. Question Number Question Text Trend Noted 4 How likely are you to refer to your current vulnerability assessment to help allocate resources to the community you serve? How likely would you support your agency budget paying for your current vulnerability assessment if it was not required by DHS/FEMA? If you do use a vulnerability assessment (like SoVI or SVI), please rate how effectively it helps you allocate resources to the community you serve from “very effective” to “not effective at all" 67% of respondents answered Likely or Very Likely of those, 32.5% were in Low Income areas and 67.5% were in Not Low Income areas 54% of respondents answered Likely or Very Likely of those, 34% were in Low Income areas and 66% were in Not Low Income areas 59% of respondents answered Effective or Very Effective of those, 26% were in Low Income areas and 74% were in Not Low Income areas 5 6 a less discriminatory national vulnerability assessment model can form. However, it is also evident from the results that poverty is a strong indicator of vulnerability and should be considered, at a minimum, as a means to test aggregated outputs. The national pilot survey called attention to several items. First, the fact that 81% of respondents answered “No” or “Do Not Know” to the second gateway question (Does your agency use a community vulnerability assessment tool like SoVI or SVI?) suggests that these indicator models 9 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 economic impact and population loss from the same or as a result of severe weather. In short, the policy could be more flexible and community focused. This will give emergency managers the flexibility, time savings, and cost to consider the best tools for their needs as opposed to the ones favored by agency policy. And while several studies like Rufat et al. [19] found that “there is a mismatch between the rising application of social vulnerability models and understanding of their empirical validity,” perhaps the emergency and disaster management community can agree that there is an increasing need to connect those vulnerability models to mitigation particularly in underserved communities. The necessary shift may begin simply with an acknowledgement of the national mitigation framework’s limitations and the consideration of SoVI because it is good enough, or the CDCs SVI because it is free and widely available. This study demonstrated that another consideration should be checking a single output of vulnerability derived from an aggregation of variables against the poverty exposure indicator which takes little or no time or cost. That irresistible temptation to jump to policy formation would require this additional criteria that represents the poorest and most vulnerable populations. Internationally, this could provide a practical example that could meet the goals of the UNDP and the Sendai Framework. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. APPENDIX. (Case study methods) 10.1 - Omitaomu and Carvalhaes [65] Indicators used in Omitaomu and Carvalhaes [65] are shown below in a table from that study. It is important to note that these authors weighted each indicator equally in an attempt to avoid subjectivity. “This subjectivity would require a tortuous justification for each weight given as a comprehensive reference for weighing these components is not yet clear” [65]. Table 2 List of the 15 ACS 2008–2012 variables used as indicators for the SVI and calculation methods to obtain final indicators to be used in the index. Colors correspond to the workflow diagram in Fig. 2. 10 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 11 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 10.2 - Martinich et al. [66] Martinich et al. [66] display the dominant variables used in their study’s Table 1 graphic below. The authors state that weighting for each regional component in their study is available upon request but mention that they employ an equal weighting of components as seen in Cutter et al. [4]. “The authors find that the manner in which the components are combined to create the final index does have a substantial impact on the end results, but they make no alternative weighting recommendation” [66]. 12 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 Table 1 Components for each regional analysis in the contiguous United States. The table displays the Social Vulnerability Index components, including the dominant variables and their associated variance explained for each regional principle components analysis North Atlantic South Atlantic Gulf Pacific Component Percent Variance Explained Component Percent Variance Explained Component Percent Variance Explained Component Percent Variance Explained Poverty Rural/Urban Age (elderly) Foreign- born Age (children) Occupation Female 22.5 11.4 10.5 8.8 7.4 5.4 5.0 Poverty Family structure Rural/Urban Foreign-born Native American/Female Occupation 20.3 17.5 11.4 10.0 6.6 4.8 Poverty Labor force participation Foreign-born Wealth Rural/Urban Occupation Female 22.4 16.6 12.3 8.2 5.6 4.4 4.0 Poverty Age (children) Rural/Urban Labor force participation Female Occupation 21.6 13.1 11.7 8.7 7.5 5.6 References [20] Tim G. Frazier, Erik Xavier Wood, Alexander Gary Peterson, Residual risk in public health and disaster management, Appl. Geogr. 125 (2020), https://doi.org/ 10.1016/j.apgeog.2020.102365, 2020,102365, ISSN 0143-6228. [21] M. Ilbeigi, S.C.K. Jagupilla, An empirical analysis of association between socioeconomic factors and communities’ exposure to natural hazards, Sustainability 12 (16) (2020) 6342. [22] S.E. Spielman, J. Tuccillo, D.C. Folch, A. Schweikert, R. Davies, N. Wood, E. Tate, Evaluating social vulnerability indicators: criteria and their application to the Social Vulnerability Index, Nat. Hazards 100 (1) (2020) 417–436. [23] J.W. Creswell, J.D. Creswell, Research Design: Qualitative, Quantitative, and Mixed Methods Approaches, fifth ed., Sage publications, 2018. [24] R. Schober, H. Annis, Barriers to help-seeking for change in drinking: a genderfocused review of the literature, Addict. Behav. 21 (1) (1996) 81–92, https://doi. org/10.1016/0306-4603(95)00041-0. [25] D. Cowan, I. Norman, V. Coopamah, Competence in nursing practice: a controversial concept – a focused review of literature, Nurse Educ. Today 25 (5) (2005) 355–362, https://doi.org/10.1016/j.nedt.2005.03.002. [26] G. Yilmaz, R. Youngreen, The application of minority influence theory in computer-mediated communication groups, Small Group Res. 47 (6) (2016) 692–719, https://doi.org/10.1177/1046496416661033. [27] F. Fatemi, A. Ardalan, B. Aguirre, N. Mansouri, I. Mohammadfam, Social vulnerability indicators in disasters: findings from a systematic review, International journal of disaster risk reduction 22 (2017) 219–227. [28] A. Shalaby, R. Tateishi, Remote sensing and GIS for mapping and monitoring land cover and land-use changes in the Northwestern coastal zone of Egypt, Appl. Geogr. 27 (1) (2007) 28–41. [29] N. Fielding, C.A. Cisneros-Puebla, CAQDAS-GIS convergence: toward a new integrated mixed method research practice? J. Mix. Methods Res. 3 (4) (2009) 349–370. [30] Payne, M. & Gerrow-Wilcox, K. Explore Maps and Scenes with the Compare Configurable App. ArcGIS Blog. Retrieved from https://www.esri.com/arcgis-blo g/products/arcgis-online/announcements/explore-maps-and-scenes-with-the-com pare-configurable-app/. [31] K. Tierney, The Social Roots of Risk: Producing Disasters, Promoting Resilience, Stanford University Press, 2014. [32] California Governor’s Office of Emergency Services (Cal OES, n.d.). County/Region Contacts. Retrieved on 9-24-2020 from https://www.caloes.ca.gov/cal-oes-divis ions/regional-operations/contacts. [33] K.A. Barchard, J. Williams, Practical advice for conducting ethical online experiments and questionnaires for United States psychologists, Behav. Res. Methods 40 (4) (2008) 1111–1128. [34] J. Losby, A. Wetmore, CDC Coffee Break: Using Likert Scales in Evaluation Survey Work, Centers for Disease Control and Prevention, 2012. [35] J.T. Croasmun, L. Ostrom, Using likert-type scales in the social sciences, J. Adult Educ. 40 (1) (2011) 19–22. [36] G. Smith, Planning for Post-disaster Recovery: A Review of the United States Disaster Assistance Framework, Public Entity Risk Institute, Fairfax, VA, 2011. [37] University of South Carolina (U of SC). (n.d.). FAQ: Social Vulnerability Index. College of Arts & Sciences Hazards & Vulnerability Research Institute. Retrieved from https://artsandsciences.sc.edu/geog/hvri/faq. [38] G2, Esri ArcGIS Pricing, 2018. Retrieved from, https://www.g2.com/products/esr i-arcgis/pricing#:~:text=The%20costs%20of%20this%20software,%2412% 2C000%20respectively%20for%20initial%20purchase. [39] National Research Council (NRC), Disaster Resilience: A National Imperative, The National Academies Press, Washington, DC, 2012, https://doi.org/10.17226/ 13457. [40] C.T. Emrich, E. Tate, S.E. Larson, Y. Zhou, Measuring social equity in flood recovery funding, Environ. Hazards 19 (3) (2020) 228–250. [41] R. Chetty, M. Stepner, S. Abraham, S. Lin, B. Scuderi, N. Turner, D. Cutler, The association between income and life expectancy in the United States, 2001-2014, Jama 315 (16) (2016) 1750–1766. [42] APM Research Labs, THE COLOR OF CORONAVIRUS: COVID-19 DEATHS BY RACE AND ETHNICITY IN THE U.S, September 16, 2020. Retrieved on 9/22/2020 from, https://www.apmresearchlab.org/covid/deaths-by-race. [1] S. Chang, J. Yip, S. van Zijll de Jong, R. Chaster, A. Lowcock, Using vulnerability indicators to develop resilience networks: a similarity approach, Nat. Hazards 78 (3) (2015) 1827–1841. [2] C. Aubrecht, S. Fuchs, C. Neuhold, Spatio-temporal aspects and dimensions in integrated disaster risk management, Nat. Hazards 68 (3) (2013) 1205–1216. [3] International Federation of the Red Cross (IFRC). (n.d.). What is vulnerability? Retrieved from https://www.ifrc.org/en/what-we-do/disaster-management/about -disasters/what-is-a-disaster/what-is-vulnerability/. [4] S.L. Cutter, B.J. Boruff, W.L. Shirley, Social vulnerability to environmental hazards, Soc. Sci. Q. 84 (2) (2003) 242–261. [5] B. Beccari, A Comparative Analysis of Disaster Risk, Vulnerability and Resilience Composite Indicators. PLoS Currents, vol. 8, Before You Apply for Building Resilient Infrastructure and Communities (BRIC) Funds |FEMA.Gov, 2016. Www. Fema.Gov, www.fema.gov/grants/mitigation/building-resilient- infrastructurecommunities/before-apply#eligibility. (Accessed 4 November 2020). [6] T.G. Frazier, C.M. Thompson, R.J. Dezzani, A framework for the development of the SERV model: a Spatially Explicit Resilience-Vulnerability model, Appl. Geogr. 51 (2014) 158–172. [7] National Center for Disaster Preparedness (NCDP), US Natural Hazards Index – National Center for Disaster Preparedness, 2021. Retrieved, https://ncdp.columbia .edu/library/mapsmapping-projects/us-natural-hazards-index/. (Accessed 12 February 2021). [8] Federal Emergency Management Agency (FEMA), FEMA Releases National Risk Index: New Online Data Shows Natural Hazards Risks For Communities, FEMA. gov, 2020 (2020). Retrieved, https://www.fema.gov/press-release/2020111 7/fema-publica-el-indice-nacional-de-riesgo-nuevos-datos-disponibles-en-linea. (Accessed 12 February 2021). [9] E. Chan, Z. Huang, H. Lam, C. Wong, Q. Zou, Health vulnerability index for disaster risk reduction: application in Belt and Road initiative (BRI) region, Int. J. Environ. Res. Publ. Health 16 (3) (2019) 380, https://doi.org/10.3390/ijerph16030380. [10] CDC’s Social Vulnerability Index (SVI). (2020, September 15). Retrieved January 24, 2021, from https://www.atsdr.cdc.gov/placeandhealth/svi/index.html. [11] Homeland Security Digital Library (HSDL). (n.d.). Robert T. Stafford Disaster Relief and Emergency Assistance Act – November 23, 1988. Retrieved from https://www. hsdl.org/?view&did=806354. [12] Federal Emergency Management Agency (FEMA). (n.d.). Mitigation Best Practices: Mapping Risks and Vulnerabilities to Increase Resilience Planning. Retrieved from https://www.fema.gov/media-library-data/1461933984986-bbb94b669d354f 1bff4420a8272f46df/20-Mapping-Risks-and-Vulnerabilities_web2.pdf. [13] B.E. Flanagan, E.W. Gregory, E.J. Hallisey, J.L. Heitgerd, B. Lewis, A social vulnerability index for disaster management, J. Homel. Secur. Emerg. Manag. 8 (1) (2011). [14] J. Evans, D. Hardy, M.E. Hauer, Social Vulnerability and Local Hazard Mitigation Planning: Application And Evaluation of a “SoVILite” Approach for Glynn County, Technical Report, Georgia, USA, 2014. [15] M. Saisana, A. Saltelli, S. Tarantola, Uncertainty and sensitivity analysis techniques as tools for the quality assessment of composite indicators, J. Roy. Stat. Soc. 168 (2) (2005) 307–323. [16] E. Tate, Social vulnerability indices: a comparative assessment using uncertainty and sensitivity analysis, Nat. Hazards 63 (2) (2012) 325–347. [17] H.A. Tarling, Comparative Analysis of Social Vulnerability Indices: CDC’s SVI and SoVI®, 2017. [18] C. Anderson, M. Hagenlocher, F. Renaud, Z. Sebesvari, S. Cutter, C. Emrich, Comparing index-based vulnerability assessments in the Mississippi Delta: implications of contrasting theories, indicators, and aggregation methodologies, International Journal of Disaster Risk Reduction 39 (2019) 101128, https://doi. org/10.1016/j.ijdrr.2019.101128. [19] S. Rufat, E. Tate, C.T. Emrich, F. Antolini, How valid are social vulnerability models? Ann. Assoc. Am. Geogr. 109 (4) (2019) 1131–1153. 13 E. Wood et al. International Journal of Disaster Risk Reduction 63 (2021) 102464 [43] J.T. Chen, N. Krieger, Revealing the unequal burden of COVID-19 by income, race/ ethnicity, and household crowding: US county versus zip code analyses, J. Publ. Health Manag. Pract. 27 (1) (2021) S43–S56. [44] S. Damaske, J.L. Bratter, A. Frech, Single mother families and employment, race, and poverty in changing economic times, Soc. Sci. Res. 62 (2017) 120–133, https://doi.org/10.1016/j.ssresearch.2016.08.008. [45] B.H. Morrow, Identifying and mapping community vulnerability, Disasters 23 (1) (1999) 1–18. [46] K. Tierney, Social Inequality, Hazards and Disasters. On Risk and Disasters: Lesson from Hurricane Katrina, RJ Daniels, DF Kettl and H. Kunreuther, 2006. [47] J. Farley, D. Baker, D. Batker, C. Koliba, R. Matteson, R. Mills, J. Pittman, Opening the policy window for ecological economics: Katrina as a focusing event, Ecol. Econ. 63 (2–3) (2007) 344–354. [48] N. Klein, The Shock Doctrine: the Rise of Disaster Capitalism, Macmillan, 2007. [49] E. Fussell, The long-term recovery of new orleans’ population after hurricane Katrina, Am. Behav. Sci. 59 (10) (2015) 1231–1245. [50] J.C. Mutter, The Disaster Profiteers: How Natural Disasters Make the Rich Richer and the Poor Even Poorer, Macmillan, 2015. [51] S. Hallegatte, A. Vogt-Schilb, M. Bangalore, J. Rozenberg, Unbreakable: Building the Resilience of the Poor in the Face of Natural Disasters, World Bank Publications, 2016. [52] D.K.P. Yee, Violence and disaster Capitalism in post-Haiyan Philippines, Peace Rev. 30 (2) (2018) 160–167. [53] E.X. Wood, T.G. Frazier, Disasters, community vulnerability, and poverty: the intersection between economics and emergency management, Journal of Emergency Management (2021), https://doi.org/10.5055/jem.0563. [54] Los Angeles County Public Health Department, LA County Daily COVID-19 Data, 2020. Retrieved on, http://publichealth.lacounty.gov/media/Coronavirus/data/ index.htm. (Accessed 2 December 2020). [55] C. Glynn, E. Fox, Dynamics OF homelessness IN urban America, Ann. Appl. Stat. 13 (1) (2019) 573–605, https://doi.org/10.1214/18-AOAS1200. [56] C. Glynn, A. Casey, Homelessness Rises Faster where Rent Exceeds a Third of Income, Zillow, Seattle, WA, USA, 2018 (Tech. Rep). [57] National Low Income Housing Coalition (NLIHC), Out Of Reach – the High Cost Of Housing Report, 2020. Retrieved from, https://reports.nlihc.org/sites/default/fil es/oor/OOR_BOOK_2020.pdf. [58] S. Hallegatte, M. Bangalore, L. Bonzanigo, M. Fay, T. Kane, U. Narloch, A. VogtSchilb, Shock Waves: Managing the Impacts of Climate Change on Poverty, The World Bank, 2015. [59] S.H. Eriksen, K. O’brien, Vulnerability, poverty and the need for sustainable adaptation measures, Clim. Pol. 7 (4) (2007) 337–352. [60] M. Gall, K.A. Borden, C.T. Emrich, S.L. Cutter, The unsustainable trend of natural hazard losses in the United States, Sustainability 3 (11) (2011) 2157–2181, https://doi.org/10.3390/su3112157. [61] Disaster Mitigation Act of 2000 (n.d.). Retrieved January 24, 2021, from htt ps://www.govinfo.gov/app/details/PLAW-106publ390. [62] Post-Katrina Emergency Management Reform Act of 2006 (PKEMRA), S.3721 109th Congress (2005-2006): Post-Katrina Emergency Management Reform Act of 2006, Congress.Gov, 2020, 2006. www.congress.gov/bill/109th-congress/senatebill/3721. [63] Social Vulnerability Assessment Tools for Climate Change and DRR Programming | UNDP Climate Change Adaptation.” Www.Adaptation-Undp.Org, www. adaptation- undp.org/Social-Vulnerability-Assessment-Tools. Accessed 4 Nov. 2020. [64] Sendai Framework for Disaster Risk Reduction 2015-2030. Available at: http ://www.wcdrr.org/preparatory/post2015. [65] O.A. Omitaomu, T.M. Carvalhaes, Developing a Climate-Induced Social Vulnerability Index for Urban Areas: A Case Study Of East Tennessee (No. ORNL/TM- 2017/353), Oak Ridge National Lab.(ORNL), Oak Ridge, TN (United States), 2017. [66] J. Martinich, J. Neumann, L. Ludwig, L. Jantarasami, Risks of sea level rise to disadvantaged communities in the United States, Mitig. Adapt. Strategies Glob. Change 18 (2) (2013) 169–185. [67] B. Flanagan, E. Hallisey, E. Adams, A. Lavery, Measuring community vulnerability to natural and anthropogenic hazards: the Centers for Disease control and prevention’s social vulnerability index, J. Environ. Health 80 (10) (2018) 34–36. [68] Gisp, C. T. E. P. D. (n.d.) Social Vulnerability and Hazard Analysis for Hurricane Harvey (Doctoral dissertation Dr. Christopher T. Emrich. University of Central Florida). [69] E.A. Lehnert, G. Wilt, B. Flanagan, E. Hallisey, Spatial exploration of the CDC’s Social Vulnerability Index and heat-related health outcomes in Georgia, International journal of disaster risk reduction 46 (2020) 101517. [70] K.H. Morin, Value of a pilot study, J. Nurs. Educ. 52 (10) (2013) 547–548. [71] J. Ivey, The value of a pilot study, Pediatr. Nurs. 40 (4) (2014) 204, 186. [72] Building Resilient Infrastructure and Communities (BRIC). (2020, September). Retrieved January 24, 2021, from https://www.fema.gov/grants/mitigation/bui lding-resilient-infrastructure-communities. [73] Guide to Expanding Mitigation: Making the Connection to Equity, Retrieved, https://www.fema.gov/sites/default/files/2020-09/fema_region-2_guide-conne cting-mitigation-equity_09-10-2020, 2020, September. (Accessed 23 January 2021). [74] C.B. Barrett, M.A. Constas, Toward a theory of resilience for international development applications, Proc. Natl. Acad. Sci. 111 (40) (2014) 14625–14630. [75] Carter C. Price, Kathryn A. Edwards, Working Papers. Trends in Income From 1975 to 2018, Rand Corporation, 2020, https://doi.org/10.7249/WRA516-1. [76] B. Wisner, P. Blaikie, T. Cannon, I. Davis. At Risk: Natural Hazards, People’s Vulnerability and Disasters, Routledge, London, 2004. 14