Portions of this article are excerpted from the author’s Cornell University Dissertation
The original dataset analysis for this article’s quantitative portion was finalized in 2021 using the latest FEMA data available – that from 2018 – which surveyed approximately 5,000 Americans. Since I first submitted this article, FEMA released two additional years of raw survey data and I was able to code the information from 2017 as well. There are now four year-groups (2017-2020) available which increased the survey population to over 20,000 people and more than doubled (from 181 to 432) the number of island resident responses. Preliminary analysis of this updated and aggregated data supports the main findings contained in this article.
Additionally, as an active-duty Colonel in the U.S. Army, I just completed a tour at the headquarters of U.S. INDOPACOM. In that assignment I was responsible for several disaster plans and orders. The experiences, exercises, observations, strategy sessions, and lessons learned from INDOPACOM align with the main themes of this paper. Typhoon Mawar, which struck Guam this year (2023) and current tensions with both North Korea and China are testaments to this. I now serve at NORTHCOM.
Residents in American island states and territories are more likely to be prepared for disaster than their mainland counterparts; anywhere from 40-134%. Island dwellers face scales of disaster that are often catastrophic for a far greater percentage of the populace than inhabitants in continentally based locales and some mitigation efforts for disasters such as hurricanes makes one more vulnerable to deadly earthquakes. Why prepare at longer levels of self-sufficiency if the chance of death or personal property destruction is greater? I analyze a new dataset from FEMA. Results indicate FEMA’s assumptions on preparedness efforts are good but lack a physical isolation variable. To attempt a remedy, I take an inference from disaster and prepper literature regarding state failure and conduct a plausibility probe. I proffer that island residents’ very vulnerability, caused by geographic realities and their real or perceived exposure to hazards that are concomitant with state failure, triggers action to increase resiliency via shelter-in-place mitigation actions.
Ellis, Chris. “Nowhere to Run to, Nowhere to Hide: Disasters, Preparedness, and the Shadow of State Failure on US Islands” Homeland Security Affairs 19, Article 3 (Sept 2023) www.hsaj.org/articles22412
“Global political economy is a field of study that deals with the interaction between political and economic forces. At its centre have always been questions of human welfare and how these might be related to state behaviour and corporate interests in different parts of the world. Despite this, major approaches in the field have often focused more on the international system perspective. A side effect of this has been the relative neglect of non-elites and an all-too-often missing recognition of ordinary individuals.” – Günter Walzenbach 
While disaster is ubiquitous to the U.S., the disparate nature is wide ranging. In this century alone, the U.S. has faced large scale terrorist attacks, a pandemic outbreak, tornadoes, hurricanes, and wildfires. Individual or household preparedness, in the face of these threats, is also varied. There are a host of ways to demarcate how to split groups for comparison but here I focus on one, the geographical difference of living on a U.S. island as compared to living on the continent. Does that single variable change the readiness level of an individual? And if so, why? Simply stated: do island residents prepare for disasters at higher, lower, or equal rates to their mainland counterparts? Due to a lack of data, this puzzle has remained unanswered and is a clear gap in the academic literature.
In the 2018 National Household Survey Executive Summary, FEMA stated five variables were causal to household disaster preparedness: awareness of information, experience with disasters, preparedness efficacy (split between two variables of belief that preparedness could help and secondly a confidence in their ability to prepare), and risk perception. Unfortunately, until late 2019 and early 2020, FEMA only released certain summary statistics on their findings. Using the 2018 released dataset, I test what I call FEMA’s Influencer Model (FIM) for explanatory power.
Living on an island has a major positive correlation (between 40-134%) on disaster preparedness and yet the FEMA model does not account for why. Adding my island variable improves the explanatory power of FEMA’s model. This article makes two contributions to the nexus of research on human security and disaster preparedness. First, by parsing out the data from FEMA it adds to the literature regarding preparedness levels at both a macro and micro level. It shows correlations of factors related to these preparedness levels and indicates that one of FEMA’s own independent variables and their overall model for preparedness do not quite match the results from their predictions for their own survey. Second, it suggests there is a potential omitted variable or collection of variables that deserve further scrutiny to solve the issue of the interaction (or better specification) among state failure, risk perception, and disaster severity. I proffer that U.S. island residents prepare at higher levels for disasters due to the likelihood of sub-national state failure, a theme that appears from both traditional disaster research and the prepper literature. I use FEMA’s data from multiple extreme catastrophes, both natural and manmade, to support this claim.
State failure comes on two related fronts. First, due to their size, islands face a higher probability of a catastrophe impacting most or all of the entire landmass. While nations can take steps to reduce or even prevent smaller disasters such as flooding, terror attacks, chemical spills, or food contamination, they cannot thwart hurricanes, earthquakes, or volcanic eruptions. Because of these limitations, islanders face systemic level impacts from just a single event, one of near totality and no – or few – safe areas to flee to for the masses. Secondly, because of the magnitude of these larger scale disasters, state capacity – and for clarity I am referring to U.S. states and territories here – is often completely overwhelmed in an analogous fashion to Small Island Developing States (SIDS) country level failures. There are large scale power outages, infrastructure damage, limited transportation avenues, and degraded communications. The sub-national state is therefore not up to the full task of disaster response. Based upon separation of hundreds, or even thousands of miles, help and recovery assets are not just physically distant, but temporally distant as well.
My research provides a potential causal explanation channel via individual action in response to this security and economic conundrum. I propose that, ceteris paribus, this state failure probability forces more islanders to confront their vulnerability and that they therefore attempt to fill this void via the gathering of additional emergency supplies to shelter-in-place. With nowhere to run to and nowhere to hide, I argue island residents are motivated to take action on emergency measures to increase the amount of time they can last while awaiting outside aid. I fill the omitted variable lacking in the FEMA model with a proxy: island or continental status.
Islands are uniquely vulnerable but can serve as observational test cases for larger catastrophes for individuals and communities in larger nations, a real-world trial run. Policymakers should take note of potential scalability when considering strategic national plans or regional, multi-state support. There are tremendous cost savings regarding preparedness and disaster planning, scalable from individual to national levels. There may also be a threshold effect at play. Existential risk, a lack of escape options, and geographic separation (including delayed response efforts) drive higher preparedness. Disaster planners should consider this when crafting public messaging as well as encouraging individual-level confidence in resilience efforts.
The article is organized as follows. In section two I link strands of human security, disaster preparedness and Small Island Developing States (SIDS). I show that most U.S. citizens are a rather low resilience group and introduce FEMA’s conjectures as a model to explain variation in their readiness efforts. Next is the SIDS information which showcases the vulnerability of these populations to larger scale disasters. I summarize the puzzle and potential answers and implications. In section three, I introduce the FEMA dataset, and test their model in a variety of manners. Section four contains an exploratory foray into a better explanation of why U.S. islanders prepare at higher levels than those on the mainland.
BACKGROUND: THE THREE STREAMS
Stream# 1: Human Security and State Failure
The United Nations Human Development Report of 1994 first introduced the term human security and tied it to calamity at inception. Since this document was written in the early years after the Cold War, the preeminent fears of a nuclear holocaust, global war between the USSR and the U.S. directly, or smaller regional wars – either proxy or not – were receding from the forefront in the discussion of international security. In this vein the document declares “[f]or most people, a feeling of insecurity arises more from worries about daily life than from the dread of a cataclysmic world event.” The report listed several sub-components of human security such as economic, environmental, and personal security. Thus, the unit of analysis switched from nations to the smallest level possible: individuals. But the relation of all these sundry “securities” was opaque.
In 2001, Roland Paris published a matrix of security studies to demonstrate the relationship between the source of the security threat and the entity of protection as depicted in Figure 1. Where his matrix is assistive here is connecting multiple threat types, all of which have bearing on individuals (earthquakes, civil war, conventional war, banking and currency collapses, etc.), and critically, showing that states cannot prevent all forms of peril. Thus, human security incorporates the range of “worries of daily life” (e.g., hunger or job opportunities) up to “cataclysmic” events with death or economic ruin as outcomes.
Figure 1: Threat and Security Matrix
Where my research furthers the debate is not on human security’s definition or development, but rather its conceptual connections to disaster preparedness. At its base, security is about the diminution of harm. A nation cannot thwart all forms of hazards from impacting people. Hence, this prevention of all harm relates to one of the definitions of state failure, which takes a broader approach to the lexicon. Rather than the term “state failure” equaling “failed state”, it can instead be described as the inability to provide basic services.
Both Zartman and Rotberg distinguish between a variety of services that states may provide, ranging from security to the rule of law, the protection of property, the right to political participation, provision of infrastructure and social services such as health and education. These services constitute a hierarchy, Rotberg argues. The provision of security is the most fundamental service states provide, in the sense that security is a condition for the provision of all other services. Rotberg also argues that failure should be seen as a continuum rather than as an either/or . . .
This understanding of state failure fits nicely with disasters of all sizes. There is a both a temporary inability of response for the sovereign or subnational state to render goods and services (power, medical assistance, rule of law) and a permanent inability to stop – i.e., provide 100% security prevention against – bigger catastrophes. Disasters of higher impact scale with greater levels (either in scope or duration) of state failure.
In summation, using Zartman and Rotberg’s framework, all nations fail, or are at risk for failure, at some point or another. By logical extension, sub-national states fail or will fail at some point or another. While both can take risk reduction measures, neither can provide total human security. Risk is still prevalent and some of that risk must be borne by the smallest unit, the individual.
Stream# 2: Disaster Preparedness
The 2015 Sendai Framework for Disaster Risk Reduction is the current baseline document for the United Nations. The definitions it provides for resilience, vulnerability, and hazard are used here. For the term disaster, the International Federation of Red Cross and Red Crescent (IFRC) defines it as:
[A] sudden, calamitous event that seriously disrupts the functioning of a community or society and causes human, material, and economic or environmental losses that exceed the community’s or society’s ability to cope using its own resources. Though often caused by nature, disasters can have human origins.
This disaster characterization complements the human security and the state failure concepts introduced previously. There is physical and economic harm, an inability to prevent, and a component of overwhelming impact.
At the time my data was collected, FEMA’s template of action was its 2018-2022 Strategic Plan. Of the three strategic goals listed, in position of primacy at number one was to “Build a Culture of Preparedness.” “Strategic Goal 1 promotes the idea that everyone should be prepared when disaster strikes” (emphasis added). This promotion of preparedness is not limited to casualty reduction, but also contains a heavy monetary element. The plan underscores the need for insurance, the increased costs of disaster, the realities of the federal fiscal environment, and most importantly, the six-to-one cost savings of pre-disaster spending .
FEMA’s second strategic goal was to “Ready the Nation for Catastrophic Disasters.” Notice the focus is not on landslides or run-of-the-mill events, but for catastrophic events. America fairs poorly in this regard. Some of the more ubiquitous issues as to why institutions fail to adequately prepare are normalcy bias, low political will, and a lack of focus on disaster preparedness with most energy instead looking at planning for disaster response rather than mitigation or planning efforts.
With these goals in mind, how do Americans fare? Over a decade of surveys from FEMA and others indicates that between 30-80% of Americans have the FEMA recommended fully stocked three-day emergency supply kit. This base level is a minimum. Other organizations go much further such as The American Red Cross which endorses a two-week home kit minimum; my analysis of FEMA’s 2018 National Household Survey (NHS) data shows that 75% do not possess this amount. Thus, a high vulnerability and low resilience level individually across the nation is the norm, not the exception. Since the FIM model fails to incorporate geographical location as a key independent variable, at least one assumption should be that U.S. island residents prepare at rates equal to their mainland counterparts.
Stream # 3– Islands: Vulnerability and Capacity
As a baseline comparison group, there are 52 designated locales, not all of which are sovereign entities, which comprise the collection of Small Island Development States (SIDS). For example, Puerto Rico, Guam, and the U.S. Virgin Islands are all SIDS, but are also territories of the United States, whereas Bahrain, Cuba, and Tonga are independent countries. Hawaii is not considered a SIDS. Although SIDS are global, the majority (23) are in the Caribbean, 20 in the Pacific Ocean, and the remaining nine scattered elsewhere.
There are multiple commonalities in the attitudes and actions of Small Island Developing States and continental nations regarding disaster planning, preparedness, insurance, and other policies. The critical dissimilarity between islands and larger countries is the increased systemic risk in islands. Systemic risk occurs “when a hazard will not only lead to negative effects in parts of the system, but also to failure of the system as a whole.” Much of this systemic risk is summed up in two components: disaster scope and physical isolation. First, based upon their smaller size and geographical composition, the proportional impact of disaster on the landmass and population is greater than that of most continental-based countries. For the same reasons, there is also a radical reduction in unaffected safe-haven locales. For example, looking at the Pacific island country of Tuvalu, topographically there does not exist an area high enough in elevation that is outside the range of a tsunami risk.
By one estimate, from 1980-2018 the top four highest death tolls from catastrophes have either been on an island or from an oceanic event which hit a coastal area. During that same time period, eight of the top ten economically costly disasters also involved either islands or oceanic/coastal events; five were hurricanes. One study from two International Monetary Fund (IMF) workers found that of the 511 disasters that struck small nations or territories (places with less than 1.5 million residents) after 1950, 324 of these occurrences were Caribbean SIDS. Figure 2 graphically depicts another IMF report which finds disasters hit small states more frequently and with a higher GDP cost, than mainland countries, especially SIDS in the Caribbean.
Figure 2: Vulnerability and Cost: Caribbean SIDS, Small, and Other States (1990-2014)
And yet another IMF working paper, this one focusing on Pacific Island Countries (PICs), indicated that during the period of 1980-2016 there was an aggregate likelihood of 34% per year of a disaster striking a PIC with 10.7% of the populace affected and an economic hit of 14.4% of GDP.
Within disaster scope, there is the further burden of choosing for which catastrophe should one prepare? Is it better to live in a wooden structure with a light roof? This is far more resistant to earthquakes but offers weak protection for hurricanes and typhoons and it poses a higher fire risk for blazes sparked by failing power lines, candles, and lamps during an earthquake. By contrast a concrete home is more fire and hurricane resilient, but if built improperly – which is common in poorer locales – becomes a collapsing sarcophagus in an earthquake.
The second unique challenge is, for populations surrounded by the ocean, fleeing and recovery is more difficult. This factor is physical isolation. Whereas residents of continentally based small states such as the Vatican, Lichtenstein, or Singapore can face countrywide threats due to their size, residents still have the option of land-based evacuation. In a worst-case scenario, they can simply walk to safety. Island residents cannot. The same logic applies to incoming aid. After Hurricane Katrina hit New Orleans in 2005 and Hurricane Harvey flooded Houston, the “Cajun Navy” – a hodge-podge of unaffiliated volunteers – swarmed the area to provide assistance above and beyond state, national, and non-governmental organization aid. While volunteers also came to Puerto Rico following Hurricane Maria, their numbers were limited by aircraft requirements. Speed and scale are impacted by mode of transport with maritime vessels being the slowest and air delivery – while speedy – is restricted by lower weight allowances, relative to ground and sea conveyance. To put this in perspective, in normal times Hawaii receives 98% of its regular imports via sea and 50% of these arrive at just one port: Honolulu Harbor. Severe damage to this single location from a natural or manmade event could cut aid inflow in half.
These two issues, disaster scope and physical isolation, create a conundrum: should an individual increase their resiliency to better negotiate the aftermath of a disaster, or are such steps futile, given the probability of destruction? In a follow-up study on Hawaiian residents after the 2018 ballistic missile false alarm, the majority receiving the information believed it to be true and thought that they and others might die. While some later took steps to increase their resiliency, for example by buying iodine pills or extra food and water, many did nothing. A sense of fatalism was captured by one respondent stating, “on this island there is nowhere to really hide.” This attitude was similar to that of Philippine residents who faced Typhoon Haiyan in 2013. Having seen numerous storms before, “[s]ome people even laughed at the local government vans that ferried people from the coast to evacuation centers inland, thinking that their concrete houses could withstand the typhoon, not knowing that it would be the storm surge that would destroy them.” The attitude of many was summed up as “If God wants to take us, then so be it!” While these two examples provide the normative mindset of a subset of individuals, there are also structural issues at play. Haiti serves as an illustration.
The 2010 Haiti earthquake killed 230,000 people in that country as compared to the 228,000 who perished in 14 separate countries from the 2004 Indian Ocean earthquake and tsunami. Figure 3 depicts the geographical swath of the Haitian quake. Geographically, over 20% of the country fell within the strong to extreme zones of shaking. Given that the majority of the populace resided in or near the capitol of Port-au-Prince, roughly 42% of the country’s citizens (3.9 out of 9 million) were hit. The Haitian earthquake displaced 1.5 million people and, as of 2019, over 32,000 were still living in camps.
Figure 3: The Geological Distribution of the Earthquake’s Severity
For means of comparison to the Haitian destruction, the next highest death tolls for earthquakes were the 2004 Kashmir earthquake in Pakistan and the 2008 Sichuan, China earthquake which killed between 70-90,000 people each. And while SIDS are the baseline group of understanding for my purposes here, even large islands are not immune. Juxtaposed against Haiti’s displacements, the 2011 Sendai earthquake and tsunami that struck Japan, led to a chain of events that cascaded into the Fukushima Daiichi nuclear disaster. 550,000 Japanese were evacuated. Ishimori found 97,000 residents were still displaced in 2017.
A Summary of the Puzzle
Are residents living in U.S. island states and territories more likely, less likely or just as likely to be prepared for disaster than their mainland counterparts? There are three possibilities. First, maybe islanders simply aggregate around the same mean of preparedness as the rest of the United States. If this is the case, there is very little puzzle to solve. The disaster preparedness literature cited supra gives ample evidence behind why people fail to concoct plans and take steps to mitigate potential future disasters. This potential is summed up in Hypothesis 1.
Hypothesis 1: Island residents’ disaster preparedness, as measured a) by the presence of a three-day emergency kit or b) in the total days they could last in their homes without power, running water, or transportation, will be equivalent to the U.S. mainland rates.
Second, because of the occurrence of previous events and/or perceived severity of future events, they prepare less for fear of losing their own lives or by the economic disincentive of anything they set aside for use after a disaster being destroyed. The SIDS data seems to indicate this as a plausible belief. This yields Hypothesis 2.
Hypothesis 2: Island residents’ disaster preparedness, as measured a) by the presence of a three-day emergency kit or b) in the total days they could last in their homes without power, running water, or transportation, will be lower than the U.S. mainland rates.
However, absence of evidence from the SIDS literature does not, by default, mean there is no mechanism for island residents preparing at higher levels. The explanatory variables of the FEMA Influencer Model (FIM) I test here may give insight.
Hypothesis 3: Island residents’ disaster preparedness, as measured a) by the presence of a three-day emergency kit or b) in the total days they could last in their homes without power, running water, or transportation, will be higher than the U.S. mainland rates.
For longer lead time disasters like hurricanes, Floridians can be issued mandatory evacuation orders and physically drive to safer destinations. Residents of Guam or Hawaii do not have that option. For these latter examples, fleeing via air is restricted by capacity and cost. However, not all disasters have a temporal notification. Florida may also serve as the better baseline of comparison rather than the average mainland scores given that Florida is the continental state most likely to face threats similar to those of island residence (namely hurricanes). While coastal counties of states such as the Carolinas or Texas may also face hurricanes, the entire state does not, so comparing all of Louisiana to Puerto Rico would provide inaccurate results. Does the FIM adequately explain this? I show below that even taking Florida as the best comparison case actually strengthens the claim that state failure is a key explanatory variable for islander rates of preparedness.
As a final point of comparison, the type of threat may come into play. The 2018 FEMA survey also included a nuclear risk component. Based upon, for example, North Korean missile fears, do residents of Guam or Hawaii fear nuclear disaster more than their mainland counterparts? FEMA data has this information for places like Guam, Honolulu, and Miami. Do residents of these areas prepare out of fear of hurricanes or nuclear fallout at statistically discernable rates? Is there a connection between the disaster type and preparedness levels?
DATA AND ANALYSIS
Research Design and the FEMA Dataset
“Good science” should be reliable (able to be reproduced), valid (accurate), and falsifiable. In late 2019, FEMA started releasing the raw numbers from their National Household Survey (NHS), beginning with their 2017 survey. While the 2017 investigation surveilled only the 50 U.S. states including Hawaii, the 2018 instrument included residents living in Puerto Rico, the U.S. Virgin Islands, and Guam. Therefore, for source information, I utilize data from the 2018 NHS. FEMA’s data is reproducible, and in fact is done so each year with minor adjustments. Because this was a national survey, it contains high external validity to non-participants, made all the greater due to replication from repeated annual iterations. For 2018 data, the results were released in January of 2020.
The 2018 NHS was conducted via phone with a live caller and individuals 18 years of age or older were polled. The population of 5003 respondents was composed of several subsets of individuals. The main group of approximately 2000 people was a random sample of all 50 U.S. states, plus Washington D.C., Guam, Puerto Rico, and the U.S. Virgin Islands. In the second subset, FEMA also surveyed an additional 3000 people in six oversample groups of roughly 500 each, sorted according to their place of residence and therefore their likely “hazard profile”: tornado, flood, hurricane, wildfire, earthquake, or nuclear event. Aggregately, these 3000 were a second random sample. The number of surveyed island residents was 181, of which 29 were from Hawaii, 34 from Guam, 106 from Puerto Rico, and 12 from the U.S. Virgin Islands. As delineated, I can therefore split respondents into two groups, those living in the continental United States and those living on an island and then subject the two groups to a battery of statistical analysis to interrogate my hypotheses.
There are a few points about the data collection and interpretation that must be fully noted. First, not all questions were asked of all groups. For example, those living in wildfire prone areas were asked specific questions regarding wildfires to which no other group was subject. Second, some questions led to follow-on queries based upon the response of the parent question. Third, many questions had as allowed responses a variation of “don’t know” or “refused to answer.” Because of these facts, unless otherwise indicated, I have stripped out all blank responses and additionally removed all “don’t know” or “refused to answer” replies and conducted calculations based solely upon clear answers. And fourth, there could be self-bias in the answers given. However, FEMA repeats this study annually and results seem to be fairly consistent. For sums and calculations, I used both Excel and R-Studio’s statistical software.
The Dependent Variables
My outcome of interest is preparedness levels in U.S. residents. Preparedness is a component of resiliency. Individuals, communities, and nations that have higher resiliency have a higher tolerance to resist disasters of larger scale and “bounce back” after a disaster occurs. This resilience is demonstrated in action by lower casualties and economic costs. Policymakers involved in boosting resilience should – and do – focus some of their efforts on household disaster preparedness levels.
FEMA queried households if they possessed the recommended three-day emergency kit. Specifically, the survey asks “Do you have enough supplies set aside in your home to get you through three days or more without power or running water and without transportation?” The answer is a dichotomous yes or no and is my first dependent variables (code = threedays2). Nearly 1100 people answered “no” to this screening question and only those who answered yes were asked the follow-up question of: “How many days do you think you could last in your home without power, running water, or transportation?”
The answer to this second question is my main dependent variable; the total number of days given (code = daystotal2). However, there are two points to note. First, those who answered “no” to the three-days screening question could have had zero, one, or two days of survivability. In order not to lose entirely these 1100 responses I assume that, on average, members in this group could last at least one day at home and therefore code them all at this level of survival in all direct comparisons after Figure 5. The second issue is on the upper bound; FEMA capped extreme responses at a maximum of 97 days. There could have been respondents who stated they were prepared for 120 days, 180 days, 365 days, or for several years but all would be coded as 97. The impact of this coding restriction should be small as only a miniscule number of respondents (n = 48) reached this level. In summation, the daystotal2 variable has a range of 1-97.
Testing Hypothesis 1 and 2
I start with my main outcome of interest: daystotal2. For coding I use the shorthand term CONUS to represent the continental states of the U.S. including Alaska. The variable Four_island encompasses the island state of Hawaii, plus Puerto Rico, Guam, and the collection of U.S. Virgin Islands. Hypothesis 1 predicts that there is no difference in preparedness rates between CONUS and Four_island. If there is a difference, Hypothesis 2 expects that, based upon the SIDS inferences, islanders will prepare less than their mainlander counterparts. The density of responses is shown in Figures 4 and 5.
Figure 4: The Range of Individual Preparedness Levels in Continental U.S. States
Figure 5: The Range of Individual Preparedness Levels in U.S. Islands and Territories
An overall look at Figures 4 and 5 shows several things. First, the distribution is not normal for either, each has a heavy positive skew. Second, it appears some individuals may be engaging in a temporal heuristic of disaster planning. There are spikes at 7, 14, 30, 60, and 90 days in both graphs. This could be due to mental rounding, planning using common calendar references (i.e., one month as an achieved goal), or because disaster goods are often sold in batches of weeks or months. There is also a spike at 97 days as was expected from FEMA’s cap at that amount. While there seems to be a greater proportion of individuals in islands with daystotal2 numbers at 30 days or more, it is hard to tell based on the sizable count difference between the two. To account for this, I log results as depicted in Figure 6.
Figure 6: Comparing Days of Preparedness, CONUS vs. Four_island
Additionally, as boxplots only display the median bar, and not the mean, I calculated the means and medians of both groups. Island residents report higher rates as displayed in Table 1. The mean islander is prepared to last over 40% longer than a mainlander (p < .001, NA’s = no answer).
Table 1: Summary Statistics, CONUS vs. Four_island
|Min.||1st Quartile||Median||Mean||3rd Quartile||Max||NA’s|
Looking at the states and territories together reveals that it is not just one island pulling the aggregate numbers of the others up. This was a concern since Puerto Rico was fresh from its Hurricane Maria experience the year prior to the survey and makes up 59% of all island respondents. Figure 7 shows the daystotal2 means of all 54 states, territories, and the District of Columbia. As is revealed, all islands are clearly on the upper half of the graph. Distance from the U.S. mainland does not appear to be related to days of preparedness. Puerto Rico is the closest, then the Virgin Islands, then Hawaii, then Guam, yet Puerto Rico has a higher total day count than Hawaii.
Figure 7: Mean Days of Preparedness by State or Territory
But what about the lower threshold query of a three-day’s worth of supplies, my first dependent variable? The data reveal that 79% of mainland residents and 86% of island residents reported meeting the three-day threshold (p < .05). Therefore, looking at either daystotal2 or the three-day results, islanders in either case possess higher rates of preparedness as measured by FEMA’s metrics. Therefore, Hypotheses 1 and 2 are rejected, Hypothesis 3 accepted. Island respondents have higher preparedness, ceteris paribus.
Expounding on Hypothesis 3
FEMA’s 2018 Executive Summary has five variables spread over four categories or “influencers”. Each of the variables match directly with a survey question. 1) Awareness of Information regards hearing in the past six months about how to prepare. Answers are a dummy yes or no variable. (code = Heard). 2) Experience with Disasters is also a yes or no question (code = Exp w/disaster). Both this and the previous variable are coded with yes being the higher number so that a positive coefficient indicates a positive relationship with the dependent variable. 3) Preparedness Efficacy has two variables; one is a belief that “preparing can help in a disaster” (code = Steps help) and the second is a confidence “in their abilities to prepare” (code = Confident). Both are ordinal variables with a Likert scale of one to five. 4) Risk Perception requires a bit more explanation. The question asks about which disaster “would have the biggest impact” where the respondent lives. 95% of respondents indicated at least one disaster. Crucially though, this question is one of severity. This is not to be confused with likelihood, which is the chance of occurrence. For comparison, the oversample groups had questions regarding the likelihood of a specified disaster – as chosen by FEMA’s hazard profile – manifesting sometime in the future. As an example, in the 500-person oversample group for flooding, 308 people (62%) indicated it would be unlikely for a flood to happen in the area in which they they live.. I return to this issue in section four of this paper. For now, I keep FEMA’s model, which measures perception of potential severity, not perception of potential likelihood. This variable is coded as a dummy (code = Risk), with zero being a response of “none” and one being a response of any type of disaster.
Results are shown in Table 2. Data interpretation is straightforward; each coefficient is equal to the predicted increase or decrease in the number of days total of survival at home. As before, I break out CONUS and Four_island since the FEMA Influencer Model does not consider geography. If these two separate regions are equivalent, then we should expect the results to be roughly the same. This is not the case.
For Model 1, risk is positive, but not significant. Awareness (Heard) is positive and slightly significant (p <.1). The rest of the variables for Model 1 are also statistically significant and the one with highest impact is Confident. Recall that this is a Likert scale variable, so every one-step increase predicts a rise of nearly three days of survival. According to the FIM predictions, if FEMA could raise individual confidence by just a single step, that alone would be the equivalent increase of the basic FEMA recommended three-day threshold.
Table 2: FIM and Total Days of Individual Preparedness
|Dependent variable: daystotal2|
|(1) CONUS||(2) Four_Island||(3) Island Dummy|
|Residual Std. Error||15.678 (df = 4155)||19.188 (df = 159)||15.821 (df = 4319)|
|F Statistic||44.534*** (df = 5; 4155)||1.770 (df = 5; 159)||39.757*** (df = 6; 4319)|
|Note:||+p < .1, *p < .05, **p < .01, ***p < 0.001|
Steps Help is, against FEMA predictions, negative (and even more so in Model 2). Interpretation for this result is a bit difficult due to the wording of the question. This question had five subcomponent parts so I do not know if mainland and island residents anchored more on one part or another. Perhaps when people living in CONUS are prompted not just to think about supplies, but also warnings and evacuation, they feel less likely to have multiple days of disaster provisions on hand since they might be able to drive away given advanced notice?
For Model 2, the risk variable is also not significant, but the coefficient is much larger than in Model 1. My assessment of this finding is twofold. First, the risk question is open ended, so people are free to think of any disaster that comes to mind. Some of those listed were wildfires, blizzards, a toxic chemical spill, power outages, earthquakes, a terrorist attack, or a meteorite striking the earth. Each of these disasters has radically different levels of severity and recovery timeframes so that may explain why this independent variable did not reach statistical significance. Second, island residents face disasters that can be existential in nature, so that could explain why their coefficient was so much higher. I expound on this finding in section four. For Model 2 only confidence was statistically significant. This is potentially due to the smaller number of valid responses for this group (n = 141). Intriguingly, the absolute value for every coefficient in Model 2 is larger than that in Model 1. Something is driving higher change, but for now what that is remains unclear.
Model 3 supports this conjecture. It includes the dummy variable (code = Islandyn) of whether or not someone resides in one of the four island groups. The positive covariate for this variable becomes the most impactful of all and indicates island residents believe they can last 3.6 days more than CONUS residents, ceteris paribus. This supports Hypothesis 3 with daystotal2 as the dependent variable.
Looking holistically, the FIM performs well, at least with the larger datasets. Adding an island dummy radically improves explanatory power on what effects disaster preparedness. Only the Steps Help variable remains anomalous.
Figure 8: Coefficient Plot of the FIM
The coefficient plot of the entire FEMA data set is shown in Figure 8 with the island dummy added to the FIM. Notice the wide dispersion of risk perception, especially in comparison to the other explanatory variables. As a reminder, individuals were asked to list any disaster they thought would be most impactful to where they lived. I correct for this in section four.
As to the second dependent variable, FEMA’s focus is not on increasing American levels of preparedness to 90 days, or even 30 or 14 days. Rather, it is on attaining the basic three-day emergency kit. Recall that FEMA described its “key influencers” as variables that would predict if an individual would “begin preparing for a future hazard”, not how in depth their preparedness would go. At this lower threshold then, as shown above, islanders are more likely to prepare at the three-days level than CONUS members (86% to 79%, respectively). Does the FIM predict this? Table 3 reveals the data.
Rerunning the regression analysis with three-days as the dependent variable (code = threedays2) yields results similar to the daystotal2 regression. The threedays2 dependent variable is dichotomous (1 = no, 2 = yes), so interpretation of the results shows predicted percent changes, not an increase in days (i.e., a coefficient of 0.07 indicates a 7% positive rise towards having a three-days of emergency supplies). What immediately jumps out from Table 3 is that Risk is .601 for islanders in Model 5, an amount 11 times as large as the equivalent coefficient in Model 4. Thus the predicted probability of possessing a three-day emergency kit increases by 60% if an islander simply believes there would be a disaster impacting where they lived as compared to an islander that thought no disaster would befall them.
No other explanatory variable comes even close across regression analysis of Models 4, 5, and 6. With this dependent variable, in Model 6 the island dummy is not statistically significant, and comparatively with other variables in Models 4-6, is rather low in impact. Introducing it does drastically mute the risk variable seen in Model 5; a finding that would have remained hidden if islanders were not broken out into their own separate category.
Table 3: FIM and a Three-day Emergency Kit
|Dependent variable: threedays2|
|(4) CONUS||(5) Four_island||(6) Island Dummy|
|Residual Std. Error||0.391 (df = 4280)||0.335 (df = 166)||0.389 (df = 4451)|
|F Statistic||53.378*** (df = 5; 4280)||3.550*** (df = 5; 166)||46.376*** (df = 6; 4451)|
|Note:||+p < .1, *p < .05, **p < .01, ***p < 0.001|
My interpretation of this stunning finding is that when islanders think of a “big” impact disaster, they conjure up something that is larger – potentially even existential in nature – which, given the data on severity from section two, is a logical deduction. Risk perception consequently is playing a radically different role in the minds of island residents.
Another unexpected finding is that personal experience with disaster is neither significant nor substantial in any of the three models in Table 3. This is strange in comparison to those two explanatory variable’s impacts against the longer scale daystotal2 outcome of interest. What could account for this? There are a few options. First, what may be at play is a threshold effect. Pierson describes mental heuristics over certain time horizons of cause and outcome. He explains that some things build over time without discernable change in the dependent variable, but then a triggering event, or threshold, occurs and much like and avalanche, the system then changes rapidly. Since the total days of preparedness was only asked of those who already acknowledged having a three-day emergency kit, perhaps that could be akin to the threshold requirement. For islanders, the fear of future catastrophe, once triggered, drives increased levels of preparedness and allows additional influences (e.g., past experience, confidence in ability) to manifest. This relates to a second possibility, one found in the prepper literature. Mills discovered in his ethnographic work with preppers that it wasn’t their personal experience with disaster that motivated them to start prepping, but rather seeing media reports of the impact of large disasters on others. Huddleston found something similar, the desire by preppers to prepare for potentialities. As a final possibility, it may be that mainland residents have an escape option mostly unavailable to their geographically isolated peers which drives down the severity fear of future events. In conclusion, Hypothesis 3 finds modest support with threedays2 as the outcome of interest and improves when adding the Island covariate. These findings seem initially dependable as the FIM is predicated on what motivates individuals to begin preparing for disaster.
To conclude this section, there are two main takeaway points. The first is that being an island resident has an impact on disaster preparedness, most likely due to perceptions of risk severity. I have proposed as a reason the more existential threat and concomitant state failure of certain disasters as well as the reduced ability to flee. Again, my speculation is that people living in the continental U.S. can drive to safety from several disasters whereas islanders cannot. Vulnerability for mainlanders can be reduced and resilience increased by the simple action of fleeing out of harm’s way rather than stockpiling a large quantity of emergency items. If there is safe haven to retreat to, why not leave? And since FEMA, state, and sometimes even national leaders often encourage this behavior (e.g., hurricane evacuation mandates), this response is highly plausible. Additionally, if by staying in place, the response efforts can reach you quickly after a disaster, why have more stored than you need?
Towards an Improved Understanding
One of the challenges listed above is that of parsing out and measuring risk perception. There is hazard variation in destruction potential, early or no warning, and duration of recovery. To control for this, I now focus specifically on just one disaster at a time and, for the purposes of teasing out if state failure (as defined earlier) may contribute to individual preparedness action, look only at heavily destructive hazards. Because of this, individuals might feel the need to mitigate both physical and financial risk at higher levels due to larger calamities.
This potentiality has been highlighted by traditional disaster scholars with the terror attacks of 9/11 and Hurricane Katrina as common examples. It also comes from researchers who look at the extremely prepared. Many preppers believe the government cannot prevent certain disasters and may even cause some of them, for example, an economic collapse due to profligate spending and debt. And after the fact, government disaster response may be heavily delayed. In certain instances, preppers believe that help might never come, a concept referred to as YOYO (You’re On Your Own). Preppers in the United Kingdom fear an inability of the state to prevent certain disasters and a lack of faith for timely disaster aid. My analysis of FEMA’s data seems to confirm this. This is a massive advancement given the dearth of quantitative data in the nascent prepper literature.
I test this from two perspectives, the first is from the point of view of the individual and their perception of what hazard will have the greatest impact on where they live; this is more a question of severity. The second is from the point of view of FEMA and their analysis is of probability of risk based upon geographic location. Five of FEMA’s six oversample groups were based more upon likelihood of a disaster at various terrestrial locales, with the sixth, a nuclear event, as an exception to likelihood and more a measure of severity. Some places could be on more than one list. For example, Dallas was included both as part of Texas with the threat of hurricanes as well as an urban area for a possible atomic detonation.
Natural Disasters (Hurricanes and Earthquakes)
I start with the individual level. FEMA asked what people thought would be the disaster most detrimental to them personally: “All areas of the country are subject to different types of disasters. Will you please name the types of disasters that would have the biggest impact where you live?” This question is identical to the one used to derive the explanatory variable Risk, but I now code it differently. For the FIM, I coded it as a dummy with zero being a response of “none” and one being a response of any type of disaster. Now, I return the coding to its original form. Respondents were not prompted with a list; the question was open ended. FEMA had eighteen preset answers for coding (snowstorm, terrorist attack, a land slide, etc.) plus “all,” “don’t know,” “refused,” “none,” and “other.”  Table 4 shows the results.
Table 4: Top Six Biggest Impact Risks by Group
| Winter storm|
|Land or mudslide|
For CONUS respondents, these match nicely with five of FEMA’s top six concerns. The top three events cluster in the 20s, then go to single digits for the last three. Storms replace nuclear events; the latter were less than 1% of people’s top disaster concern. Island residents show something far different. 81% said their top fear was a hurricane or earthquake. This matches well with the SIDS data. All other concerns are much lower.
Isolating only those citizens who indicated a hurricane as their top fear, I created a new subgroup of individuals and then once again split out CONUS and Four_island residents. Figure 9 shows the side-by-side comparison of these two and their logged days of preparedness levels.
Figure 9: Personal Beliefs on Hurricane Impacts and Preparedness, CONUS versus Islands
While the medians are the same as Figure 6, the means are 8.7 for mainlanders and 14.8 days for islanders (p < .001). This is a larger split (6.1 days or a 70% mean increase CONUS to island) than that of the full sample when testing all FEMA respondents and all hazards earlier (see Table 1 and recall that difference was 4.4 days).
Why a larger discrepancy? Possibly because hurricanes provide earlier warning, mainland residents can flee, usually by car. Islanders lack this capacity. And, since hurricane pathway forecasts are predictions, not absolutes, mainland residents can take a wait-and-see approach up until the final hours. In contrast islanders must purchase a comparatively expensive airline ticket, typically days in advance, for a disaster that may pass them altogether. Also, availability is limited both due to the number of flights and the fact that fleeing tourists might be taking up many of the seats. Moreover, mainland state governors can order mandatory evacuations. Although their island counterparts can do the same, it is not nearly as feasible. Where would the people go other than a community shelter?
For comparison, I also isolated the responses of Floridians. Of the 1057 respondents who indicated hurricane as most impactful, 339 were from Florida and 260 provided a response to their total days of preparedness. That average was 10.1 days, higher than the mainland response but much lower than that of the average island resident. When looking back at Figure 7, Florida was far below any of the island responses and 16th lowest of all states and territories. I find this to be powerful evidence that the ability to evacuate plays a heavy role in risk perception and the need or benefits of having emergency supplies on hand, and to a greater extent, for islanders versus continental citizens.
FEMA’s point of view supports the same conclusion. FEMA conducted an oversample of 500 residents living in areas FEMA deemed most susceptible to hurricanes. FEMA asked these residents how likely a hurricane was to strike their area. 411 thought it was likely and 66 thought it was unlikely. The mean total days of preparedness for CONUS was 10.7 days and Four_island was 14.9, roughly a four-day difference (p < 0.1). This result provides a robustness check and supports my contention that islanders, when controlling for the type of disaster either by perception (individual) or by probability (as computed by FEMA), are more likely to prepare for disaster at significantly higher rates than their continental brethren. Figures 10 and 11 depict the state and territory averages. It is evident that islands are on the high ends of preparedness from both metrics.
Figure 10: Mean Total Days of Individual Preparedness, Hurricane “biggest impact” Respondents Only
Figure 11: Mean Total Days of Individual Preparedness, Hurricane Oversample Respondents Only
What about earthquakes? Haiti showed the utter desolation on an island state for this type of disaster. Although, a major difference as compared to hurricanes is that earthquakes rarely come with a warning. There is no chance to flee prior to the event for either island or mainland residents. But, as indicated earlier, response and recovery efforts are easier when state, sovereign, and volunteer assets can drive to the area of the emergency.
From the FEMA data, 1124 people believed that earthquakes would be the most impactful disaster to befall their area. The daystotal2 boxplot in Figure 12 depicts the same recurring pattern. Island residents had a mean of 15.6 days of preparedness and CONUS members were 10.1 (p < .01). Figure 13 indicates the averages for respondent risk perception by residence; these are not as strong as those seen elsewhere, but earthquakes threaten a far larger swath of America than hurricanes do.
Figure 12: Personal Beliefs on Earthquake Impacts and Preparedness, CONUS versus Islands
Figure 13: Mean Total Days of Individual Preparedness, Earthquake “biggest impact” Respondents Only
When moving to FEMA’s point of view, their earthquake oversample had only nine islanders. Therefore, I cannot conduct a robustness check via this method for earthquakes.
Approaching the puzzle of why island residents prepare at higher levels than their mainland counterparts from the viewpoint of risk perception coupled with state failure seems to be on the right track. Whether it is the inability to flee, the destructive nature of certain catastrophes, the lag time in response efforts, or a combination of these items. Something is activated in islanders to gather more supplies. I conduct one final inquiry to see if this holds for a catastrophic manmade incident, that of a nuclear event. Since I have not discussed in detail a nuclear threat previously, I give below a small background of a single incident that recently happened to provide salience for the topic.
Man-made Disaster (Nuclear Event)
On January 13, 2018, all cell phones in Hawaii, including both residents and tourists received the following alert message: “BALLISTIC MISSILE THREAT INBOUND TO HAWAII. SEEK IMMEDIATE SHELTER. THIS IS NOT A DRILL.”  The alert was not without merit, escalating verbal spats including threats of nuclear use between the United States and North Korea were ongoing between the two countries. North Korea had explicitly threatened Guam in August 2017 after launching over Japan a Hwasong-12 intermediate range missile. In the months prior to the alert, the Hawaiian government had unequivocally warned the populace via television commercials to prepared for a possible nuclear attack and sounded a test of its air raid sirens on December 1st, 2017. The Hawaiian island of O’ahu is a strategic military target containing Marine Corps Base Hawaii, the Naval and Air Force Joint Base Pearl Harbor – Hickam, the Army’s Schofield Barracks, and the four-star headquarters of the United States Indo-Pacific Command, among others. Fortunately, the alert message was not real, but rather part of an exercise that a single government worker misinterpreted.
To glean information on individual actions in light of this kind of hazard, FEMA’s nuclear oversample queried residents of large urban centers (primary nuclear targets) and their suburbs such as Washington D.C., Denver, New York, Miami, and Los Angeles. FEMA also included Honolulu and residents of Guam. Responses were aggregated by state and territory. Figure 14 shows the variation of daystotal2 preparedness, by location based exclusively on FEMA’s nuclear oversample population (n = 500). Guam and Hawaii have the highest rates out of thirteen locales. There is no discernable pattern among the continental states.
Figure 14: Mean Total Days of Individual Preparedness, Nuclear Oversample Respondents Only
The boxplot, Figure 15, comparing CONUS to, in this case just two islands, shows the spread of preparedness (logged). Showcasing the summary statistics here seems helpful so they are depicted in Table 5 (NA’s = no answer).
Figure 15: Comparing Days of Individual Preparedness for a Nuclear Event, CONUS vs. Islands
Table 5: Nuclear Event daystotal2 Oversample Summary Statistics, CONUS vs. Island
|Min.||1st Quartile||Median||Mean||3rd Quartile||Max||NA’s|
There is over a full week’s difference (a 134% increase, p < 0.01) in total days of preparedness, the largest spread of any combination in all the disasters I analyzed. 175 of the 500 members (35%) of the nuclear oversample queried thought a nuclear explosion was likely where they lived. 56% of islanders thought it likely (25 of 45) and 33% of mainland residents (150 of 455).
Preparedness steps, even for a nuclear explosion, are beneficial and thus logical. While those individuals at ground zero or within the blast and overpressure radius have a high chance of immediate death or significant radiation poisoning, there is a far larger portion of the populace in the fallout zone. Each day of sheltering-in-place allows the half-life of radioactive nuclides and debris to decay to lower levels, facilitating less risky evacuation.
From the examples given in this section, it appears that both severity and likelihood are causal factors which the island variable helps bring to light. Thus, the risk perception influencer that FIM uses might not be the best explanatory variable in its current form. The interaction of perceptions on severity and likelihood, coupled with beliefs on what government can do for mitigation and response (i.e., the heuristic “breakpoint” for state failure) does seem to hold. Connection to prepper research seems initially to be validated by these findings.
Throughout a battery of hazards and statistical techniques, a consistent theme emerges that U.S. island residents prepare for disaster, via means of personally procuring supplies, at significantly higher rates than their mainland counterparts. Results were robust to both natural and manmade events as well as two separate levels, that of a three-day emergency kit and that of the total days one could stay at home without power, water, or transportation. While the FEMA Influencer Model could not fully account for the geographic split, endogenous to the risk perception might be issues of evacuation options and the fear of state failure. If true, then FEMA’s data seems to support quantitatively the prepping world’s qualitative literature supposition of the latter factor. It appears that as perception of risk severity goes up, so does the belief that official help may be long in coming. Mitigation steps to remain sheltered-in-place for longer durations act as a form of household insurance. The analysis is also in keeping with the current field of human security. As such, I focused on the preparedness of individuals rather than state action. And finally, understanding motivations with hard data contributes heavily to the burgeoning analysis of disaster preparedness at the household level.
Several pathways for future study emerge from all of this. First, there may be some concern regarding equifinality. Not all people are motivated to prepare for the same reasons. FEMA’s 2018 NHS Executive Summary hints at up to three additional frameworks of coding that could be worked into multivariate analysis. Second, FEMA’s own questionnaire could be updated to tease out some of the nuances. Regarding the fear of state failure, questions on the duration of a disaster’s impact until power, running water, or transportation would be restored would be an excellent addition. A generic version could be “How long would you expect it to take until life returned to normal?” or “How soon would you expect government aid to your household (or the community) if (disaster X, Y, or Z) happened?” Parsing out both of these against most likely and most severe disasters for a community would also be assistive, as well as the impacts of repeated exposure rather than single exposure. Mills and Huddleston both found that repeated exposure of news media coverage regarding the disasters that befell others pushed many preppers into action or kept them fueled for vigilance. Simplifying the “preparedness steps help” question could reduce uncertainty in responses from that query and coding income on a continuous scale would be some of my further recommendations. Third, national level preparedness surveys for SIDS would be a fruitful endeavor to investigate to discover if residents of Grenada prepared at levels similar to those in Guam or Puerto Rico.
Unfortunately, with such a small pool of island residents, I did not feel confident in adding additional variables such as income, race, an urban-rural divide, and several others. Due to the lack of significance achieved looking at the island only models, I feel this is prudent. Additionally, since this is a quantitative study, I may have missed factors only revealed by qualitative research. Anecdotally, I know several coastal and island residents believe that if a disaster struck, they could live off the abundant sea life or tropical fruit common in certain areas; much like hunters in rural areas think similar thoughts.
As a final note, FEMA’s Influencer Model may indirectly yield a final benefit. If citizen confidence is truly a large factor in preparing, and as history has shown there are disasters whose impact lasts more than three days, FEMA could best support people by providing templates of emergency kits of longer duration (i.e., 7-, 14-, or even 30-days) and encouraging their confidence in survival, writ large. This advice seems especially poignant given the recent crises of COVID-19 and the nuclear war discussions surrounding China and Russia. If there is truly nowhere to run to and nowhere to hide from a global calamity, increased individual resiliency is a sound and timely investment.
About the Author
Dr. Chris Ellis is an active-duty Colonel in the U.S. Army currently serving as a disaster cell director at NORTHCOM. He has participated in several large-scale disaster exercises and planning events over his career as well as five operational deployments. His work is featured in The Homeland Security Digital Library, The Atlantic Council, The Strategy Bridge, and Army Magazine. He habitually guest lectures and served on the 2020 Harvard University Pandemic Policython as a mentor. Dr. Ellis earned his undergraduate degree in Biology from the University of Washington and holds master’s degrees in: Public Administration (University of Kansas), Military Art and Science (both from the Command and General Staff College and the School of Advanced Military Studies), and Government (Cornell University). He earned his doctorate from Cornell University in Government. You can reach him at email@example.com, or follow him on Twitter @Prep4Disasters.
Aerts, Jeroen C. J. H., W. J. Wouter Botzen, Kerry Emanuel, Ning Lin, Hans de Moel, and Erwann O. Michel-Kerjan. “Evaluating Flood Resilience Strategies for Coastal Megacities.” Science 344, no. 6183 (May 2, 2014): 473–75. https://doi.org/10.1126/science.1248222.
American Red Cross. “Survival Kit Supplies.” Accessed March 20, 2020. https://www.redcross.org/get-help/how-to-prepare-for-emergencies/survival-kit-supplies.html.
Barker, Kezia. “How to Survive the End of the Future: Preppers, Pathology, and the Everyday Crisis of Insecurity.” Transactions of the Institute of British Geographers n/a, no. n/a (December 2, 2019). https://doi.org/10.1111/tran.12362.
Berke, Philip, Galen Newman, Jaekyung Lee, Tabitha Combs, Carl Kolosna, and David Salvesen. “Evaluation of Networks of Plans and Vulnerability to Hazards and Climate Change: A Resilience Scorecard.” Journal of the American Planning Association 81, no. 4 (October 2, 2015): 287–302. https://doi.org/10.1080/01944363.2015.1093954.
Charles, Jacqueline. “Ten Years After Haiti’s Earthquake: A Decade of Aftershocks and Unkept Promises.” Pulitzer Center, January 8, 2020. https://pulitzercenter.org/reporting/ten-years-after-haitis-earthquake-decade-aftershocks-and-unkept-promises.
Clarke, Lee. Worst Cases: Terror and Catastrophe in the Popular Imagination. Chicago: The University of Chicago Press, 2006.
Comfort, Louise, Arjen Boin, and Chris Demchack, eds. Designing Resilience: Preparing for Extreme Events. Pittsburgh, Pennsylvania: University of Pittsburgh Press, 2010.
Deitchman, Scott, Cham E. Dallas, and Frederick Burkle. “Lessons from Hawaii: A Blessing in Disguise.” Health Security 16, no. 3 (March 20, 2018): 213–15. https://doi.org/10.1089/hs.2018.0014.
DeYoung, Sarah E., Jeannette N. Sutton, Ashley K. Farmer, David Neal, and Katherine A. Nichols. “Death Was Not in the Agenda for the Day: Emotions, Behavioral Reactions, and Perceptions in Response to the 2018 Hawaii Wireless Emergency Alert.” International Journal of Disaster Risk Reduction 36 (May 1, 2019): 101078. https://doi.org/10.1016/j.ijdrr.2019.101078.
Duffin, Erin, and Munich Re. “Biggest Natural Disasters Worldwide by Economic Damage up to 2018.” Statista, April 29, 2019. http://www.statista.com/statistics/268126/biggest-natural-disasters-by-economic-damage-since-1980/.
Duffin, Erin, and Worldatlas.com. “Most Significant Natural Disasters Worldwide by Death Toll up to 2018.” Statista, April 29, 2019. http://www.statista.com/statistics/268029/natural-disasters-by-death-toll-since-1980/.
Dunbar, Paula, Kelly Stroker, and Heather McCullough. “Do the 2010 Haiti and Chile Earthquakes and Tsunamis Indicate Increasing Trends?” Geomatics, Natural Hazards and Risk 1, no. 2 (June 1, 2010): 95–114. https://doi.org/10.1080/19475705.2010.487322.
Ellis, Chris. Are You Ready for It? Examining Security in Contemporary Disaster Preparedness, From Normal to Noahchian. Proquest, 2021. https://t.co/RqgohGZStV.
———. “Personal Disaster Preparedness Levels in the National Guard.” Monograph, School of Advanced Military Studies, 2016.
Eriksen, Stein Sundstøl. “‘State Failure’ in Theory and Practice: The Idea of the State and the Contradictions of State Formation.” Review of International Studies 37, no. 1 (January 2011): 229–47. https://doi.org/10.1017/S0260210510000409.
Federal Emergency Management Agency. “Be Prepared for a Nuclear Explosion.” FEMA, March 2018. https://www.fema.gov/media-library-data/1522436486111-24f5ba6d06cc07500cddd4d70c53b6dc/NuclearExplosion_508.pdf.
Flynn, Stephen. The Edge of Disaster. New York: Random House, 2007.
Garrett, Bradley. Bunker: Building for the End Times. New York, NY: Scribner, 2020.
Gilens, Martin, and Benjamin I. Page. “Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens.” Perspectives on Politics 12, no. 3 (September 2014): 564–81. https://doi.org/10.1017/S1537592714001595.
Herrera, Yoshiko. “Theory, Data, and Formulation: The Unusual Case of David Laitin.” Qualitative Methods 4, no. 1 (Spring 2006).
Huddleston, Chad. Responses to Disasters and Climate Change. Chapter 22: “Prepper” as Resilient Citizen: What Preppers Can Teach Us about Surviving Disasters. Edited by Michele Companion and Miriam Chaiken. Boca Raton, Florida: CRC Press, 2016. https://www.researchgate.net/publication/315957699_Prepper_as_resilient_citizen_what_preppers_can_teach_us_about_surviving_disasters.
International Federation of Red Cross and Red Crescent. “What Is a Disaster?” What is a disaster. Accessed March 30, 2020. https://www.ifrc.org/en/what-we-do/disaster-management/about-disasters/what-is-a-disaster/.
Ishimori, Miki. “Right to Housing after Fukushima Nuclear Disaster (Working Paper).” Department of Law and Criminology of Aberystwyth University, October 31, 2017. https://www.ifrc.org/PageFiles/125745/Right%20to%20housing%20after%20Fukushima%20nuclear%20disaster.pdf.
Kaleem, Jaweed. “As North Korean Threat Grows, Hawaii Prepares for Nuclear Attack.” Los Angeles Times, November 10, 2017. https://www.latimes.com/nation/la-na-hawaii-nuke-2017-story.html.
Kang, Cecilia. “Hawaii Missile Alert Wasn’t Accidental, Officials Say, Blaming Worker.” The New York Times, January 30, 2018, sec. Technology. https://www.nytimes.com/2018/01/30/technology/fcc-hawaii-missile-alert.html.
King, Gary, Robert Keohane, and Sidney Verba. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton, New Jersey: Princeton University Press, 1994.
Kunreuther, Howard, and Erwann Michel-Kerjan. “People Get Ready: Disaster Preparedness.” Issues in Science and Technology 28, no. 1 (2011): 39–50.
Kunreuther, Howard, and Michael Useem. Learning from Catastrophes: Strategies for Reaction and Response. New Jersey: Wharton School Publishing, 2010.
Lee, Dongyeol, Huan Zhang, and Chau Nguyen. “The Economic Impact of Natural Disasters in Pacific Island Countries: Adaptation and Preparedness.” International Monetary Fund, May 2018. https://www.imf.org/en/Publications/WP/Issues/2018/05/10/The-Economic-Impact-of-Natural-Disasters-in-Pacific-Island-Countries-Adaptation-and-45826.
Lendon, Brad, and Joshua Berlinger. “Next Target Guam, North Korea Says.” CNN, August 30, 2017. https://www.cnn.com/2017/08/29/asia/north-korea-missile-launch-guam-threat/index.html.
McGlinchey (Editor), Stephen. International Relations. Bristol, England: E-International Relations Publishing, 2017.
Mills, Michael F. “Obamageddon: Fear, the Far Right, and the Rise of ‘Doomsday’ Prepping in Obama’s America.” Journal of American Studies, June 13, 2019, 1–30. https://doi.org/10.1017/S0021875819000501.
———. “Preparing for the Unknown… Unknowns: ‘Doomsday’ Prepping and Disaster Risk Anxiety in the United States.” Journal of Risk Research 22, no. 10 (October 3, 2019): 1267–79. https://doi.org/10.1080/13669877.2018.1466825.
Muir-Wood, Robert. The Cure for Catastrophe: How We Can Stop Manufacturing Natural Disasters. New York: Basic Books, 2016.
Muñoz, Sònia, and İnci Ötker. “Building Resilience to Natural Disasters in the Caribbean Requires Greater Preparedness.” IMF News, December 18, 2018. https://www.imf.org/en/News/Articles/2018/12/07/NA120718-Building-Resilience-to-Natural-Disasters-in-Caribbean-Requires-Greater-Preparedness.
Nagourney, Adam, David E. Sanger, and Johanna Barr. “Hawaii Panics After Alert About Incoming Missile Is Sent in Error.” The New York Times, January 13, 2018, sec. U.S. https://www.nytimes.com/2018/01/13/us/hawaii-missile.html.
Ötker, İnci, and Krishna Srinivasan. “Bracing for the Storm: For the Caribbean, Building Resilience Is a Matter of Survival.” Finance and Development 55, no. 1 (March 2018): 4.
Pallardy, Richard, and Encyclopedia Britannica. “Haiti Earthquake of 2010.” Encyclopedia Britannica. Accessed March 17, 2020. https://www.britannica.com/event/2010-Haiti-earthquake.
Pierson, Paul. “Big, Slow-Moving, and . . . Invisible: Macrosocial Processes in the Study of Comparative Politics.” In Comparative Historical Analysis in the Social Sciences, edited by James Mahoney and Dietrich Rueschemeyer. Cambridge, United Kingdom: Cambridge University Press, 2003.
Ponce de Leon, Maria Inez Angela Z. “The Limits of a Disaster Imagination: A Study of Two Communities Hit by Haiyan.” Journal of Risk Research, November 6, 2019, 1–15. https://doi.org/10.1080/13669877.2019.1687576.
Prizzia, Ross. “Coordinating Disaster Prevention and Management in Hawaii.” Disaster Prevention and Management: An International Journal 15, no. 2 (January 1, 2006): 275–85. https://doi.org/10.1108/09653560610659829.
Roland Paris. “Human Security: Paradigm Shift or Hot Air?” International Security 26, no. 2 (2001): 87.
Schweizer, Pia-Johanna. “Systemic Risks – Concepts and Challenges for Risk Governance.” Journal of Risk Research 0, no. 0 (November 11, 2019): 1–16. https://doi.org/10.1080/13669877.2019.1687574.
Shultz, James M., Madeline A. Cohen, Sabrina Hermosilla, Zelde Espinel, and Andrew McLean. “Disaster Risk Reduction and Sustainable Development for Small Island Developing States.” Disaster Health 3, no. 1 (January 2, 2016): 32–44. https://doi.org/10.1080/21665044.2016.1173443.
Taupo, Tauisi Minute. “A Survey of Disaster Risk and Resilience in Small Island States.” IntechOpen, November 5, 2018. https://doi.org/10.5772/intechopen.80266.
Taylor, Alan. “10 Years Since the Devastating 2008 Sichuan Earthquake.” The Atlantic, May 9, 2018. https://www.theatlantic.com/photo/2018/05/10-years-since-the-devastating-2008-sichuan-earthquake/560066/.
Tierney, Kathleen. Disasters: A Sociological Approach. Medford, MA: Polity Press, 2019.
United Nations Development Programme, ed. Human Development Report 1994. New York: Oxford Univiversity Press, 1994.
United Nations Office for Disaster Risk Reduction. “Sendai Framework for Disaster Risk Reduction 2015 – 2030.” United Nations Office for Disaster Risk Reduction, 2015.
U.S. Census Bureau. “2010 Earthquake in Haiti.” The United States Census Bureau, July 8, 2019. https://www.census.gov/topics/preparedness/events/earthquakes/haiti.html.
US Federal Emergency Management Agency. “2018 National Household Survey Results.” Executive Summary. Washington D.C.: FEMA, January 2, 2020. https://www.fema.gov/media-library/assets/documents/182373.
———. “Strategic Plan (2018-2022).” Government Printing Office, November 30, 2018. https://www.fema.gov/strategic-plan.
 As quoted in McGlinchey (Editor), International Relations..
 US Federal Emergency Management Agency, “2018 National Household Survey Results.”
 Ellis, Are You Ready for It? Examining Security in Contemporary Disaster Preparedness, From Normal to Noahchian; Flynn, The Edge of Disaster.
 United Nations Development Programme, Human Development Report 1994.
 Roland Paris, “Human Security.”
 Eriksen, “‘State Failure’ in Theory and Practice.”
 United Nations Office for Disaster Risk Reduction, “Sendai Framework for Disaster Risk Reduction 2015 – 2030.”
 International Federation of Red Cross and Red Crescent, “What Is a Disaster?”
 US Federal Emergency Management Agency, “Strategic Plan (2018-2022).”
 Flynn, The Edge of Disaster; Kunreuther and Michel-Kerjan, “People Get Ready”; Kunreuther and Useem, Learning from Catastrophes: Strategies for Reaction and Response.
 Ellis, “Personal Disaster Preparedness Levels in the National Guard”; US Federal Emergency Management Agency, “2018 National Household Survey Results.”
 American Red Cross, “Survival Kit Supplies.”
 See for example Aerts et al., “Evaluating Flood Resilience Strategies for Coastal Megacities”; Berke et al., “Evaluation of Networks of Plans and Vulnerability to Hazards and Climate Change.”
 Schweizer, “Systemic Risks – Concepts and Challenges for Risk Governance.”
 Shultz et al., “Disaster Risk Reduction and Sustainable Development for Small Island Developing States.”
 Taupo, “A Survey of Disaster Risk and Resilience in Small Island States.”
 Duffin and Worldatlas.com, “Most Significant Natural Disasters Worldwide by Death Toll up to 2018.”
 Duffin and Munich Re., “Biggest Natural Disasters Worldwide by Economic Damage up to 2018.”
 Ötker and Srinivasan, “Bracing for the Storm: For the Caribbean, Building Resilience Is a Matter of Survival.”
 Muñoz and Ötker, “Building Resilience to Natural Disasters in the Caribbean Requires Greater Preparedness.”
 Lee, Zhang, and Nguyen, “The Economic Impact of Natural Disasters in Pacific Island Countries.”
 Muir-Wood, The Cure for Catastrophe: How We Can Stop Manufacturing Natural Disasters.
 Shultz et al., “Disaster Risk Reduction and Sustainable Development for Small Island Developing States.”
 Prizzia, “Coordinating Disaster Prevention and Management in Hawaii.”
 DeYoung et al., “‘Death Was Not in the Agenda for the Day.’”
 Ponce de Leon, “The Limits of a Disaster Imagination.”
 Dunbar, Stroker, and McCullough, “Do the 2010 Haiti and Chile Earthquakes and Tsunamis Indicate Increasing Trends?”
 U.S. Census Bureau, “2010 Earthquake in Haiti.”
 Charles, “Ten Years After Haiti’s Earthquake.”
 Pallardy and Encyclopedia Britannica, “Haiti Earthquake of 2010.”
 Duffin and Worldatlas.com, “Most Significant Natural Disasters Worldwide by Death Toll up to 2018”; Taylor, “10 Years Since the Devastating 2008 Sichuan Earthquake.”
 Ishimori, “Right to Housing after Fukushima Nuclear Disaster (Working Paper).”
 King, Keohane, and Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research.
 See Author’s Preface.
 US Federal Emergency Management Agency, “2018 National Household Survey Results.”
 Many questions were in regards to the overall household’s level of readiness. Therefore, I often interchange the term “individual” with “household.”
 While there may be some individuals who truly could not last one day at home, I assume that populace is outweighed by those who could last two days. Thus, if anything, my assumption underestimates total days of survival.
 Due to the size of the y-axis, several low count days (less than five responses) cannot be seen in this figure.
 See for example https://www.costco.com/emergency-kits-supplies.html, https://beprepared.com/, https://www.samsclub.com/b/emergency-food-storage-kits/1760103, https://beprepared.com/, or https://mypatriotsupply.com/collections/emergency-survival-food.
 Awareness of Information: “In the past six months, have you read, seen, or heard any information about how to get better prepared for a disaster? By disaster, I mean events that could threaten lives, disrupt public or emergency services like water and power, or damage property.”
Experience with Disasters: “Have you or your family ever experienced the impacts of a disaster?”
Preparedness Efficacy. 1) Preparedness steps help: “How much would taking steps to prepare, such as creating a household emergency plan, developing an evacuation and shelter plan, signing up for alerts and warning systems, or stocking up on supplies help you get through a disaster in your area?”2) Confident: “How confident are you that you can take the steps to prepare for a disaster in your area?”
Risk Perception: All areas of the country are subject to different types of disasters. Will you please name the types of disasters that would have the biggest impact where you live?
 Responses such as “don’t know” and “refused” are treated as NA. This applies to all of the variables I indicate as coded.
 Respondents were allowed to give any answer they wanted. 18 specific disasters were coded, plus “other, all, none, don’t know, and refused.” Respondents were also allowed to choose up to three answers. In FEMA’s 2018 Executive Summary, they state 98% of respondents “acknowledge that the occurrence of at least one disaster type could impact where they live” but this is not technically true. My direct correspondence with FEMA indicate they took the ~2% who answered “none” and treated all others as a positive. I correct for this in my coding and analysis.
 Table Two and Table Three are both Ordinary Least Squares regression.
 There were two questions from the FEMA survey that also queried on evacuation. The results indicate islanders are slightly more likely to have an evacuation plan and have emergency supplies packed and ready to grab immediately if needed. This further supports my claim that islanders think about disasters more than mainland residents.
 Pierson, “Big, Slow-Moving, and . . . Invisible: Macrosocial Processes in the Study of Comparative Politics.”
 Mills, “Preparing for the Unknown… Unknowns.”
 Huddleston, “Prepper” as Resilient Citizen.
 Clarke, Worst Cases: Terror and Catastrophe in the Popular Imagination; Comfort, Boin, and Demchack, Designing Resilience: Preparing for Extreme Events.
 Tierney, Disasters: A Sociological Approach.
 Mills, “Obamageddon.”
 Mills, “Preparing for the Unknown… Unknowns.”
 Garrett, Bunker: Building for the End Times; Huddleston, “Prepper” as Resilient Citizen.
 Barker, “How to Survive the End of the Future.”
 The “other” category had synonyms for certain hazards, i.e., typhoons instead of hurricanes. I manually counted several of these to add to Table 4. All data in this section besides Table 4 though comes strictly from FEMA’s 19 preset answer responses.
 Snowstorms, Ice storms, Extreme Winter Weather, Extreme Heat, and Flooding were also preset FEMA responses. These events can sometimes have a multi-day warning prior to occurring. Comparing CONUS to Island responses shows islanders with a 2.5-day tally advantage.
 I ran the regressions with the FIM and the added Islandyn variable and found the latter to be statistically significant and the largest covariate both in the individual severity grouping of data and in the FEMA likelihood grouping.
 Caution is warranted regarding individual states for data interpretation throughout this section. Sample sizes are often extremely small. Here it is best to look only at generalized regional comparisons. More raw data is needed.
 Nagourney, Sanger, and Barr, “Hawaii Panics After Alert About Incoming Missile Is Sent in Error.”
 Lendon and Berlinger, “Next Target Guam, North Korea Says.”
 Kaleem, “As North Korean Threat Grows, Hawaii Prepares for Nuclear Attack.”
 Kang, “Hawaii Missile Alert Wasn’t Accidental, Officials Say, Blaming Worker.”
 Deitchman, Dallas, and Burkle, “Lessons from Hawaii.”
 Deitchman, Dallas, and Burkle; Federal Emergency Management Agency, “Be Prepared for a Nuclear Explosion.”
Copyright © 2023 by the author(s). Homeland Security Affairs is an academic journal available free of charge to individuals and institutions. Because the purpose of this publication is the widest possible dissemination of knowledge, copies of this journal and the articles contained herein may be printed or downloaded and redistributed for personal, research or educational purposes free of charge and without permission. Any commercial use of Homeland Security Affairs or the articles published herein is expressly prohibited without the written consent of the copyright holder. The copyright of all articles published in Homeland Security Affairs rests with the author(s) of the article. Homeland Security Affairs is the online journal of the Naval Postgraduate School Center for Homeland Defense and Security (CHDS).