Many kinds of emergencies can lead to a necessary evacuation. In some cases, people may have a day or 2 to prepare while other situations may call for an immediate evacuation. 1 Planning is essential to making sure that households can evacuate quickly and safely no matter what the circumstances. 1 Encouraging residents in areas with higher risk of hurricanes to evacuate before a storm makes landfall, for instance, is one way to reduce hurricane-related morbidity and mortality. However, many factors have been shown to discourage a household’s decision to evacuate. According to recent surveys by the Centers for Disease Control and Prevention (CDC), when respondents were surveyed for evacuation barriers, the primary reasons they were unable to evacuate were because of pet concerns, cost, and if community COVID-19 levels in their area were high. 2 In light of this problem, it is important for public health interventions to increase evacuation rates, especially among groups at increased risk for negative outcomes (e.g., older adults, persons with mobility challenges, persons who are experiencing homelessness).Reference Ricchetti-Masterson and Horney 3
Previous research has shown that those who experienced a past hurricane without major harm, those who believe their home is not in a community placed at increased risk for hurricane damage, and those that think the storm is not severe enough to warrant evacuation, are less likely to evacuate.Reference Ricchetti-Masterson and Horney 3 Most residents who feel unsafe staying where they are during a storm tend to leave, and those who feel safe tend to stay.Reference Ricchetti-Masterson and Horney 3 When Hurricane Ida hit New York City in 2021, flooding occurred rapidly in basement apartments during the night—when decedents were home—with little time to evacuate.Reference Yuan, Spira-Cohen and Olson 4 Records indicate how quickly flooding overtook people and, in several cases, decedents were actively trying to evacuate and could not escape. In at least 1 case, the decedent was sleeping and deaths occurred before the state of emergency declaration, which also lacked basement-specific safety messaging; many decedents spoke limited English.Reference Yuan, Spira-Cohen and Olson 4 The circumstances of these Ida-related deaths illustrate the importance of issuing early weather and basement-specific warnings in multiple languages, as well as installing basement flood alarms to mitigate risk for residents who are sleeping or have not yet noticed signs of flooding (if any).Reference Yuan, Spira-Cohen and Olson 4
Literature shows that people expected to evacuate often do not end up doing so and those who should not evacuate often do.Reference Bartolucci and Lüders 5 People do not always respond to risk warnings as authorities hope, primarily because they feel as if their overall options are limited.Reference Bartolucci and Lüders 5 Once an evacuation order is put into place, people decide on whether to evacuate or not based on previous experiences and available information at hand.Reference Bartolucci and Lüders 5 Research showed that some of the most important factors that influence decision-making in the face of disaster include socioeconomic variables such as the presence of children or family members; demographic indicators such as gender, race, and ethnicity; personal experiences with previous disaster; property ownership; and social ties with friends and family.Reference Bartolucci and Lüders 5 Continued research on evacuation behaviors is needed overall, particularly on how timing affects evacuation. A better understanding may help to reduce barriers and improve evacuation compliance. Planning and preparing communities for hurricanes and other natural disasters can be stressful and complex, more so during the COVID-19 pandemic. 6 To understand how the COVID-19 pandemic may affect preparedness during disasters, in June 2020 the CDC surveyed a sample of 500 adults from across the country. The survey asked respondents how the pandemic may affect their plans to shelter for disasters, including hurricanes, tornadoes, and wildfires. 6 Respondents mentioned that concerns about getting COVID-19 could keep them from going to a shelter during an extreme weather incident, and that they would not be able to frequently wash their hands. CDC explored these concerns further with an online survey of 3000 adults in 8 states along the Atlantic and Gulf Coasts in October 2020. Respondents said they had changed their emergency response plans because of the COVID-19 pandemic and listed fears about going to a shelter, such as other people not wearing masks, being unable to keep distance from those outside their households, and concern about older family members getting COVID-19. 6 These data show how the current climate can affect preparedness and response behaviors among households. To further expand understanding and increase overall knowledge of evacuation behaviors, the goal of this manuscript is to characterize the potential natural disaster evacuation behaviors amidst the COVID-19 pandemic among a nationally representative sample.
Methods
To understand potential evacuation behaviors during the COVID-19 pandemic, data were collected and analyzed through Porter Novelli’s (PN) ConsumerStyles surveys. PN ConsumerStyles are cross-sectional market surveys of a random sample of non-institutionalized adults (aged 18 years or older) from Ipsos’ KnowledgePanel®. To address self-selection bias, panel members must be invited to join through random recruitment by mail using probability-based sampling by address to reach respondents regardless of whether they have landline phones or Internet access. Approximately 5 weeks after the initial mailing, telephone refusal-conversation calls are made to nonresponding households for which a telephone number is matched. If needed, households are provided with a laptop or tablet and a mobile data plan for Internet access as all surveys are online only.
In 2020, FallStyles was sent to a sample (n = 4548) of panelists, fielded from September 14, /2020 to October 10, 2020, who answered the 2020 SpringStyles survey, fielded from March 19, 2020 to April 9, 2020. In 2021, SpringStyles, was sent to 10 919 panelists between March 23, 2021 and April 13, 2021 with 6455 adults (59.1%) completing the survey. For the 2020 FallStyles survey, reminders were sent to non-responders on days 3, 7, and 13, and those who completed the survey received 5000 cash-equivalent reward points and were eligible for a sweepstakes. For 2021 SpringStyles, email reminders were sent to all non-responders on day 3, 6, 9 of the field period. Three additional reminders were sent on days 13, 16, and 19 in order to maximize response rates. Those who completed the survey also received 5000 cash-equivalent reward points (worth approximately $5).
The Fall 2021 wave, FallStyles, was fielded from September 24, 2021 to October 7, 2021. The survey was sent to a sample of 4510 panelists who answered the SpringStyles 2021 survey. Email reminders were sent to non-responders on days 3, 7, and 11 of the field periods. Survey completion time was approximately 37 minutes (median). Respondents were not required to answer any of the questions and could exit the survey at any time. Those who completed the survey received 10 000 cash-equivalent reward points (worth approximately $10). Respondents who did not answer at least half of the questions or completed the survey in 10 minutes or less were removed from the data as incomplete (n = 31). A total of 3553 adults completed the survey for a response rate of 78.8%. While sampled from the same KnowledgePanel® pool, the 2020 FallStyles, 2021 SpringStyles, and 2021 FallStyles are separate samples; there is no way of knowing if any respondents participated in all surveys.
While the specific questions related to evacuation remained the same in all surveys, there were changes to some demographic variables between Fall 2020 and Fall 2021. All modifications were accounted for by creating matching variables between 2020 FallStyles, 2021 SpringStyles, and 2021 FallStyles, except for employment which could not be aligned and is noted as such in the tables. In addition, the 2020 FallStyles survey included an additional question on potential barriers to going to a shelter during COVID-19 not included in SpringStyles or FallStyles 2021. We conducted weighted analysis of the data using SAS version 9.4 to examine distributions and estimate associations of potential evacuation behaviors of each survey separately. FallStyles 2020 and 2021 weights are based off the previous SpringStyles and adjusted according to the US Current Population Survey (CPS) proportions, while SpringStyles data were weighted using the 2019 Census’ American Community Survey (ACS) proportions.
Descriptive analysis, using means and frequencies, were used to examine distributions of demographic characteristics and potential evacuation behaviors. Missing data were minimal in all surveys for all variables (< 5%). Chi-square tests were used to examine the association between evacuation behaviors and demographics, disaster experience, perceptions of preparedness, emergency supply kits, and disaster risk. Because FallStyles and SpringStyles data were similar in terms of descriptive statistics and significant associations, multivariable logistic regression was run on FallStyles 2021 only to help explain the relationship between key variables (e.g., race, ethnicity, income, education) and evacuation behaviors.
Multivariable logistic regression, using the backward stepwise elimination procedures, were used beginning with all variables in the model (either all demographic factors, all disaster experience variables, or all beliefs) and eliminating those that did not statistically predict (P < 0.05) the dependent variable (evacuation behaviors) one by one. Only the final model is presented in the text. All data presented within this report, including the tables, are weighted. Data are presented with Fall 2020 first, followed by Spring 2021, and then Fall 2021 unless otherwise noted. However, data are presented as one value if they were the same for the 3 surveys. If the 3 data points had less than 1% difference, they are reported as one value with an approximate (~) sign.
Results
Descriptive Results
Overall, the weighted demographics were comparable across the 3 surveys (Table 1). Slightly more than half of respondents (51.6%) were female. Roughly 63% self-identified as White with ∼11% Black, ∼16% Hispanic, and less than 2% mixed-raced (Table 1). Most live in single-family homes (73.1%, 71.7%, 72.2%), with ∼15% in apartment homes, ∼8% in townhomes or duplexes, and ∼4% in mobile homes, RVs, boats, or vans (Table 1). The majority (73.7%, 72.5%, 72.7%) own their homes with a quarter (24.4%, 25.6%, 25.7%) renting and ~2% living in their home without payment (Table 1). The South had the most representation of respondents with ∼38%, followed by the West (24%), Midwest (∼21%), and Northeast (∼17%), with the majority living in metro areas (86.6%) compared to non-metro (13.4%). Less than 15% live alone (Table 1).
Table 1. Weighted demographics of respondents
* Fall 2020 “Employed” includes all currently employed persons, and “Other” includes those who are temporarily out of work; Spring 2021 and Fall 2021 “Employed” is employed full-time only, and “Other” are those who are employed part-time. Therefore, these are separate categories and should not be compared.
Most respondents (69.0%, 63.5%, 55.5%) have experienced a disaster, with severe weather with power outages being the most common (55.1%, 50.3%, 40.9%) followed by a tropical storm or hurricane (29.2%, 23.4%, 22.9%) (Table 2). Respondents also indicated experiencing the following disasters: a tornado (15.7%, 13.8%, 10.2%); earthquake, mudslide, or landslide (15.5%, 14.2%, 10.9%); or flood (14.2%, 12.2%, 11.7%) (Table 2). Respondents (16.4%, 19%, 16.1%) reported that they, or somebody in their household, worked, volunteered, or trained in disaster response or recovery (Table 2).
Table 2. Weighted preparedness levels and disaster experience
When asked about barriers to evacuation if public authorities announced a mandatory evacuation because of a large-scale disaster, ~57% reported there would be nothing that would prevent them from evacuating (Table 2). However, (20.6%, 19.7%, 21.0%) reported a concern of leaving pets, (21.9%, 19.0%, 20.3%) were concerned about leaving their property, and roughly 12% to 15% said they had nowhere to go (Table 2). Few (5.5%, 5.3%, 5.6%) cited health problems or a lack of transportation (3.8%, 3.8%, 4.3%) as a barrier (Table 2).
Overall, less than 3% had all 5 FEMA preparedness plans (Table 2). Throughout all 3 surveys, 56%-57% responded that they felt confident they knew how to prepare for a disaster (Table 3). Evacuation was highest among those with a bachelor’s degree and those 75+ years and older (Table 4). When looking at evacuation behaviors by preparedness, disaster experience and beliefs were considered, among those who had an emergency supply kit (ESK), (58.8%, 61.3%, 59.1%) evacuated. When considering preparedness plans, (67.7%, 64.6%, 61.9%) of those who had all plans evacuated. Among those that experienced a previous disaster (severe weather with outages, hurricane/storm, tornado, earthquake/landslide, flood, and wildfires) 56.3%, 56.4%, 54.6% evacuated. When looking at individuals who were confident on how to prepare for a disaster, 59.4%, 60.4%, and 58.1% of them evacuated (Table 5). In the chi-square analysis for all 3 cycles, there is a significant association found between evacuation behavior and age, education, race/ethnicity, housing structure, housing status, urbanicity, marital status, and household income (Table 4). In addition, there is a statistically significant association between evacuation behavior and preparedness levels as well as emergency supply kit cost (Table 5).
Table 3. Weighted beliefs about disasters and preparedness
Table 4. Evacuation behavior by demographic characteristics
* Fall 2020 “Employed” includes all currently employed persons, and “Other” includes those who are temporarily out of work; Spring 2021 and Fall 2021 “Employed” is employed full-time only, and “Other” are those who are employed part-time. Therefore, these are separate categories and should not be compared.
Table 5. Evacuation behaviors by preparedness, disaster experience and beliefs
Multivariable Regression Results
Age, education level, race/ethnicity, housing structure, region, and household income all remained significant predictors of preparedness in the final adjusted model (Table 6). Those aged 75+ years had an increased odds (OR 2.6; 95% CI 1.9, 3.5) of evacuation, followed by those aged 55-75 years (OR, 1.7; 95% CI 1.5, 2.1) and 35-54 years (OR, 1.3; 95% CI 1.1, 1.6) compared to the reference age group of those aged 18-34 years. Non-Hispanic Blacks and Hispanics had increased odds of reporting they would evacuate (OR 1.8; 95% CI 1.4, 2.2 and OR 1.2; 95% CI 1.0, 1.5, respectively) compared to non-Hispanic Whites. Although not statistically significant, mixed-race respondents had a decreased odds of evacuation (OR, 0.7; 95% CI 0.4, 1.1). Having an income of $50,000 or higher was associated with a significantly higher odds of evacuating when compared to those with an income of less than $25,000. Those in the Midwest (OR 1.2; 95% CI 1.0, 1.5) or West (OR 1.3; 95% CI 1.1, 1.6) were significantly more likely to evacuate than those in the South (Table 6).
Table 6. Weighted logistic regression analysis of evacuation behavior by demographics factors, Fall 2021
Discussion
Assessing the preparedness levels of communities can help public health and other agencies plan for disasters or emergencies and tailor messaging to increase community preparedness. This can include determining if households have emergency plans; supply kits with adequate food, water, and medicine; preferred and trusted communication sources; and, intended evacuation (or non-evacuation) plans.Reference Schnall, Wolkin and Roth 7 Knowledge of such potential evacuation plans and behaviors is a key aspect to understand for preparedness for any disaster or emergency. By understanding the potential evacuation behaviors of the community, local jurisdictions can tailor messages and communication campaigns to the community regarding evacuation, including developing interventions to improve disaster preparedness and response behaviors.
Based on these data, only about 60% of people across the nation would evacuate if told to do so. Messaging on the importance of evacuation is essential to help people safely navigate through the impacts of disasters. Understanding the hesitancies of the remaining 40% allows for overall barriers to be addressed on a local as well as national level. Top concerns of those reporting they would not evacuate were leaving pets and property. This is comparable with data from several Community Assessments for Public Health Emergency Response (CASPERs) conducted prior to the COVID-19 pandemic and is important to recognize as a consistent barrier to safe evacuation behaviors.Reference Schnall, Wolkin and Roth 7 – 10 One potential explanation could be a lack of awareness of shelter locations, safety measures, and/or the availability of pet-friendly shelters. It is important to work with representatives on a local and national level to increase communications on overall access to shelters in various communities.
On top of these more traditional barriers, the Fall 2020 data also demonstrated specific pandemic-related concerns about evacuation during an incident such as concerns about getting COVID-19, lack of social distancing, inadequate sanitizing, and medical care access. This is similar to other survey data from the pandemic. 11 Although there is no longer a disaster declaration for COVID-19, people may still have hesitancies about the spread of infectious diseases in shelters, hotels, or other congregate settings, which supports the need for continued messaging about safety precautions being taken. In addition, infectious diseases are an important consideration that public health and emergency management should continue to address when planning evacuation shelters.
Importantly, almost 15% of respondents reported nowhere to go as a barrier, despite shelters being available for free and often accounting for concerns such as pets by providing a safe location at or nearby the shelter for pets. In addition, while lack of transportation was cited by less than 5% of respondents, this is an important issue to address, especially when understanding overall barriers during evacuation. Overall, transportation is an important factor during evacuation to ensure communities can evacuate in a timely manner. It is essential for public health to work with emergency management to ensure that all those who want to evacuate can safely do so, whether it is through better communication about the availability of safe sheltering, providing transportation to those in need, or developing creative solutions for those who are concerned about leaving their property behind.
Several individual factors (e.g., age group, race/ethnicity, perceptions and beliefs, previous disaster experience) indicate the importance of messaging to be specifically directed to how people can prepare emergency plans prior to a disaster occurring. It may be possible that they do not have the resources or information at hand before a disaster occurs to prepare accordingly. Our data show significant relationships between evacuation and age, education, ethnicity, housing structure, household status, region, urbanicity, marital status, household income, and employment status. This has implications in preparedness and messaging, and future efforts should tailor interventions to specific demographic groups in hopes to increase preparedness and evacuation. Tailored messaging and engaging with the local communities are essential for communities to have a plan in the time of emergency. It allows for those in disaster-prone areas to evacuate in a timely manner prior to a disaster occurring.
When looking at the preparedness levels overall, many respondents would have benefited if there were an increase of preparedness plans overall, along with an increase in confidence on how to prepare for a disaster to be able to safely navigate through and prevent burden from these disasters. While studies have shown that local media communication is a critical source of information regarding disasters, and thus emergency preparedness, changes in messaging may be necessary overall. 12 These data serve to highlight this need on a national scale, but they would also be helpful in assisting with the customization and tailoring of certain initiatives at the local level. Agencies and organizations at all levels—federal, state, local, tribal, and territorial (STLT), community—can comprehend the gaps and set the right priorities for their population of focus by doing so. Because disasters happen locally, preparatory measures also need to be locally developed and implemented. The skills and relationships of local trusted leaders (e.g., religious leaders, local personalities, popular business owners, community organizers) and community members should be leveraged to focus on specific groups. These trusted individuals can help change social norms and understanding about preparedness and encourage planning.Reference Schnall, Kieszak and Heiman 13 The first step in closing the household preparedness gap is recognizing the variations in a community’s beliefs, potential risks, and demographics.Reference Schnall, Kieszak and Heiman 13
Survey findings indicate that perceived risk and confidence were significant factors in readiness. There was a lower likelihood of preparedness among those who felt that the risk of contracting an infectious disease outweighed the risk of a disaster. In terms of focusing messages and communication efforts to homes, this is crucial to understand. If someone is confident in their disaster preparedness or does not believe that a disaster will affect their home, they may not follow the communication guidance provided to them. This has also been observed with pandemic flu when the public is informed of the risks; subjective risk assessment affects the level of preparedness, and therefore, education strategies must take into account expectations, social context, and the influence and trust of the health agency. This is also consistent with theories suggesting that communication must be adapted to different stages of thinking/belief and overall disaster preparedness.Reference Schnall, Kieszak and Heiman 13 Disasters can, and do, happen everywhere and to anybody. The benefits of avoiding the threat, and factors influencing the decision to act, can support changes in perception that help advance behavior changes for increased preparedness.Reference Schnall, Kieszak and Heiman 13
Limitations
These data are not without limitations. ConsumerStyles surveys are cross-sectional and limited to only those within the panel. Therefore, while there are 3 surveys, they are only 3 snapshots in time and do not represent a longitudinal analysis. Also, even though KnowledgePanel® works to ensure representativeness of the respondents on several key aspects, there are some potential differences in areas that have traditionally mattered in disaster preparedness and response, such as household structure, home ownership, and persons within the home (e.g., marital status, living with others, having kids). Further, the panel only represents those within the 50 US states and does not include panel members from US territories. The US territories are prone to disasters and should be included in all disaster research. As far as the survey questions, the demographic categories changed between Fall 2020, Spring 2021, and Fall 2021, making it impossible to compare employment and limiting the analysis of household type by combining mobile homes with boats, RVs, and vans. A final limitation noted is that the overall response regarding evacuation is hypothetical—whether the respondent thinks they would evacuate or not—rather than actual evacuation. Additionally, because all questions were closed-ended, any reasoning for certain responses (e.g., “other”) had to be inferred. While this research is integral in acquiring knowledge of evacuation behaviors, a needed step is to explore in detail with more granular data. Overall, developing policies that help raise awareness of the need and opportunity for preparedness actions and plans at the household-level would help increase preparedness on a broader scope. Efforts must continue to be made at the local level to inform and address preparedness. In addition, these data can be valuable for informing community outreach and engagement and the tailoring of sometimes limited resources. These data can also help inform response planning and the updating of communication resources such as websites, fact sheets, and other materials to reach a wide audience.Reference Schnall, Kieszak and Heiman 13 In addition, understanding how evacuation can decrease morbidity and mortality should be studied further.
Conclusion
Overall, these data show that there is much work to be done in terms of evacuation behaviors and overall preparedness in the United States. These data are an essential starting point in determining evacuation behaviors to help tailor public messaging, work with partners to increase knowledge in evacuation, and guide future research. Therefore, efforts must continue to be made at the local level to both inform and address evacuation preparedness. These include focused communication strategies to address barriers, including those related to preparedness and planning. It is important to understand how the data can be implemented for various audiences and specifically depending on the environment in which they live.
Disclaimer
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Author contribution
Sumera Jiva: Main report writing; Arianna Hanchey: Contributed to report writing; Stephanie Kieszak: Statistical analysis, methodology, contributed to report writing; Amy Schnall: Concept development, contributed to report writing, methodology.
