Disaster Preparedness and Housing Tenure: How Do Subsidized Renters Fare?

Abstract

Homeowners are significantly more prepared for disasters than renters. However, disaster preparedness among subsidized renters is an understudied topic despite their increased vulnerability to negative disaster outcomes. Previous research shows that one in three subsidized units is at risk for exposure to disasters, relative to one in four unsubsidized rental units, and one in seven owner-occupied units. Subsidized housing residents often fall into many vulnerable statuses that would make them less prepared than renters and owners. Using 2017 American Housing Survey data, we examine differences in household disaster preparedness by housing tenure, with and without controls for such factors. Logistic regression analyses indicate that subsidized renters are significantly less prepared than unsubsidized renters, and both renter types are significantly less prepared than homeowners, controlling for demographic and neighborhood characteristics. The policy implications of this research are considered as they relate to the location and management of subsidized housing in an era of climate change.

Introduction

From 1980 to 2021, the United States experienced 308 severe weather and climate disasters with damages that reached or exceeded one-billion dollars in cost (NOAA 2021). In 2021 alone, 688 people died as the result of these costly hazardous events (NOAA 2022). An estimated 6 million occupied rental units, which compose about 14% of the occupied rental housing stock, are located in areas with a relatively or very high risk of experiencing annual losses from climate events (JCHS 2022). With regards to subsidized housing, a much higher share are located in these vulnerable areas; specifically, 1.2 million out of the total 3.1 million Low-Income Housing Tax Credit (LIHTC) units (39%), 700,000 out of the total 2.3 million project -based HUD units (30%), and 200,000 out of the total 400,000 USDA multifamily units (50%) are located in high-risk areas (JCHS 2022).

Nationally, over 200,000 federally assisted units are in a regulatory floodway or within a 100-year flood plain, and an additional 242,000 subsidized units are in a 500-year flood plain (PAHRC and NLIHC 2021). Flooding can reduce the supply of federally subsidized housing units in a county for up to three years after the event because it takes that much time, on average, for the old units to be repaired and new units to be constructed (Davlasheridze and Miao 2021). Flooding can also increase subsidized housing waitlist times and the amount subsidized tenants pay in rent due to a reduction of available units in the short term (Davlasheridze and Miao 2021). It is estimated that the number of affordable units, both subsidized and not, at risk for flooding will more than triple by 2050 (Buchanan et al. 2020).

Despite this disproportionate risk to disaster among households with housing subsidies, there is little research on disaster preparedness among subsidized housing residents in the U.S. (Khajehei and Chandrasekhar 2021; Murphy et al. 2009; PAHRC & NLIHC 2021; Rivera 2020). A 2021 joint report by the Public and Affordable Housing Research Corporation (PAHRC) & the National Low Income Housing Coalition (NLIHC) found that when compared to other low-income renters, federally subsidized renters were significantly less likely than unassisted low-income renters to have access to a vehicle or evacuation funds of $2,000. These differences in preparedness are likely due to limited financial resources and health limitations. However, these studies do not control for factors like age, race and ethnicity, education, or disability status when examining housing tenure.

The primary research goals of this paper are twofold. The first is to examine whether subsidized renters differ in their disaster preparedness from unsubsidized renters and homeowners. The second is to explore whether differences persist after controlling for variation in demographic and socioeconomic characteristics. We use 2017 American Housing Survey (AHS) data to address these research goals. These data classify subsidized renters as those receiving HUD assistance and living in public housing or privately-owned multifamily housing, or participants in the Housing Choice Voucher (HCV) program. While prior research shows that homeowners are more prepared for disasters than renters (Murphy et al. 2009; Rivera 2020), there is minimal research comparing disaster preparation differences between different types of renters, relative to homeowners (Khajehei and Chandrasekhar 2021; PAHRC & NLIHC 2021). In this study we define household disaster preparedness as a combination of resource-based (i.e., stockpiling food and water; preparing an emergency supply kit; having: evacuation funds, property or renter’s insurance, access to a vehicle, CO detectors, and/or an electric generator) and action-based (i.e., having: access to financial information and contact numbers; a communications plan; and/or established evacuation meeting location) outcomes, which we discuss more fully below in the data and methods section.

To preview our findings, first, we find that differences in household disaster preparedness exist for those living in subsidized housing, relative to unsubsidized renters and homeowners. Second, we examine differences in preparedness and account for them by considering household demographic and socioeconomic characteristics that have not been examined thus far; we find that many significant differences between subsidized renters and unsubsidized renters and homeowners remain with the inclusion of these control variables. Finally, we conclude that policy measures are needed to bolster the preparedness of households in subsidized units as well suggest that disaster risk be considered in future research on housing inequality.

Background

Disasters are not “natural” (Squires and Hartman 2013). There are predictable patterns in who is affected and how severely they are affected by extreme climate events, thereby making disasters social in nature (Evans 2010). Social vulnerability refers to a population’s sensitivity to natural hazards as well its ability to respond and recover from hazards (Cutter and Finch 2007). Vulnerabilities to natural hazards are created by social structures and manifest as poverty, food insecurity, poor housing stock, and environmental degradation. Race and ethnicity, socioeconomic status, gender, and age significantly shape who is able to respond and recover from climate events (Cutter and Finch 2007; Wood, Burton, and Cutter 2010). Thus, disaster events are a combination of social vulnerabilities and natural hazards (Evans 2010; Wang and Yarnal 2012). Social vulnerability and exposure to disasters also apply to hazards that do not center around weather hazards, such as nuclear risk and acts of terrorism (Boyd and Scouras 2010; Cutter, Richardson, and Wilbanks 2003; Eisenman et al. 2006; Fekete 2022; Zhukova 2020). Given the nature of social vulnerability to hazardous events, households must have resource-based and action-oriented plans in place to prepare themselves for future emergencies (Zamboni and Martin 2020).

The literature on disaster preparedness focuses mostly on the socioeconomic and demographic characteristics of individuals and households and how these characteristics relate to their ability to prepare for severe climate events and other non-climate and weather-related hazards (Kohn et al. 2012). There is also attention paid to the perceived and actual risk of experiencing a severe climate event and how both factors shape preparedness (Bell et al. 2020; Murphy et al. 2009; Nukpezah 2020). Recently, a third category of correlates has been examined – those relating to how residential location relates to a household’s ability to prepare (Burby et al. 2003; Friedman et al. 2021; Lee and Van Zandt 2019; Murphy et al. 2009; Rivera 2020).

Correlates of Preparedness and Housing Tenure

Numerous studies on disaster preparedness reveal that income, educational attainment, race and ethnicity, age, and disability are all indictors of how likely a household is to prepare for natural hazards (Cox and Kim 2018; DeBastiani et al. 2015; Friedman et al. 2021; Kohn et al. 2012; Zamboni and Martin 2020). Lower-income households have fewer financial resources to prepare and are thus less prepared overall than households with higher incomes (Al-rousan, Rubenstein, and Wallace 2014; Cox and Kim 2018; Crowley 2021; Kohn et al. 2012). Educational attainment is also positively associated with preparedness (DeBastiani et al. 2015; Zamboni and Martin 2020).

Black and Hispanic Americans are, in general, not only more vulnerable to disasters in the U.S. (Crowley 2021; Levine et al. 2007), but also more likely to experience precarious housing conditions (Díaz McConnell 2017). However, there are inconsistent findings in the literature regarding how race and ethnicity influences disaster preparedness (Kohn et al. 2012). Some studies yield no significant association while others indicate that Black and Hispanic households are less prepared than White households (Al-rousan et al. 2014; Murphy et al. 2009). These different findings may stem from the ways race and ethnicity, educational attainment, and income work together to influence disaster preparedness. Additionally, examining the ways in which White, Hispanic, and Black households prepare yields different findings. For example, White households often gather material items and financial resources while Black and Hispanic households create emergency plans or stockpile items that require minimal financial resources (Kim and Zakour 2017; Zamboni and Martin 2020).

Age is also a nuanced predictor of preparedness. While adults 30 years of age or older are more prepared than younger adults, adults aged 65 and older are the most likely to prepared (Kohn et al. 2021). However, the oldest among older adults (that is adults 85 and older) are less prepared than older adults aged 65 to 74 years (Al-rousan et al.2014; Cox and Kim 2018). Poor health among the oldest adults is a factor that reduces their disaster preparedness (Shapira et al. 2019). Collectively, this indicates that the relationship between age and preparedness is an inverted U-shape, with preparedness being lowest in the younger ages, peaking in middle age, and declining among older adults (Heller 2005).

While differences between men and women are not often the focus of disaster preparedness research, overall men tend to have higher levels of preparedness than women (DeBastinani and Strine 2013; Gazibara, Haomiao, and Lubetkin 2014). Women-headed households are especially underprepared when it comes to resource-based items (Zamboni and Martin 2020).

People with disabilities are also more vulnerable to disasters and are less prepared than people without disabilities (Al-rousan et al. 2014). They often have compounding disadvantages such as having less education and more economic-, housing-, food-, and employment-based insecurities; in addition, community-disaster planners often fail to factor in their needs (Fox et al. 2007; Sabatello et al. 2020). However, households with a person with a disability are more likely to have an electric generator than those without disabled persons (Zamboni and Martin 2020). Notably disaster preparedness among adults younger than 65-years-old with disabilities is understudied. Much of the current literature examines disabilities as one factor that makes older adults vulnerable (Al-rousan et al. 2014; Cox and Kim 2018). About 22% of the 60 million Americans living in poverty have a disability and are thus even more vulnerable to disaster (U.S. Department of Homeland Security 2020).

Subsidized renters are disproportionately represented among the demographic groups that are least likely to prepare, relative to unsubsidized renters and homeowners. Subsidized renters living in federally assisted properties often have more financial challenges than unsubsidized renters and thus often lack the funds to prepare and for temporary shelter during and after evacuation (PAHRC and NLIHC 2021). For example, in 2017, the average household yearly income for HUD subsidized renters was $13,958 (HUD 2017). Comparatively, according to the American Community Survey (ACS), the median income for all renters was $38,944, and the homeowner median yearly income in 2017 was $78,876 (U.S. Census Bureau 2017a). With respect to education, in 2017, 21.5% of all HUD-assisted householders had no high school diploma compared to 10.6% of all renters, and only 8.3% of HUD-assisted householders had a bachelor’s degree or higher, relative to 15.2% of all renters (HUD 2022a). Meanwhile, only 11.5% of homeowners lack a high school diploma, and 32.9% of owners have a bachelor’s degree or higher (U.S. Census Bureau 2017b).

Black householders are overrepresented among subsidized renters, composing 45.5% of all HUD-assisted renters compared to 21.4% of all renters. Black households only represent 8.0% of homeowners (U.S. Census Bureau 2017b). In 2017, the proportion of subsidized renters that were Hispanic, however, was slightly less at 18.4% than all renters at 19.9% (HUD 2022a). Among homeowners in 2017, only 9.3% were Hispanic (U.S. Census Bureau 2017b). Among subsidized renter households, 27.5% include occupants that are 65-years-of-age or older compared to 14.6% of all renters and 30.2% of owner-occupied units (HUD 2022a; U.S. Census Bureau 2017b). Finally, 43.6% of subsidized households have at least one member with a person with a disability while only 22.3% of all renter households have a person with a disability (HUD 2022a).

A second set of factors that correlate with preparedness include the actual and perceived risk of experiencing a disaster (Friedman et al. 2021; Kievik and Gutteling 2011; Malmin 2021; Martins et al. 2019; Rivera 2020). Past experience with disasters is positively associated with stockpiling medication, relocating vehicles, securing furniture, and participating in disaster drills (Han et al. 2017; Tomio, Sato, and Mizumura 2012). These findings hold true among subsidized renters as well. Two years after Hurricane Sandy, renters living in units owned and operated by PHAs in New York City who did not evacuate reported taking steps to prepare for future disasters by stockpiling supplies of food, water, batteries, and flashlights as well as developing family evacuation plans (Hernández et al. 2018).¹

Research on the correlation between perceived risk of disasters and preparedness is more mixed. Some studies find that perceived risk of disaster is positively associated with preparedness (Basolo, Steinberg, and Grant 2017; Friedman et al. 2021; Martins et al. 2019). In addition, households that perceive they are more in control and able to handle a disaster are more likely to prepare (Kievik and Gutteling 2011; Reininger et al. 2013). However, perceived risk of disaster as well as feeling a lack of control can result in households failing to prepare (Han et al. 2017). Based upon these findings, it remains unclear if previous disaster experience and perception of risk will differ by housing tenure. Subsidized renters are more vulnerable to disasters so they might be more likely to prepare than unsubsidized renters and owners. Alternatively, homeowners have more of a vested interest in protecting their housing, and their perceptions of risk of disaster are likely to be associated with their preparedness, perhaps more than among subsidized and unsubsidized renters.

Residential characteristics are a third set of factors associated with household disaster preparedness. Most studies that include residential characteristics focus on housing tenure as a key characteristic in being associated with preparedness, although none of those studies disaggregate renters into subsidized and unsubsidized renters (Burby, Steinberg, and Basolo 2003; Murphy et al. 2009). Studies comparing renters to homeowners find that homeowners tend to be better prepared than renters (Laska and Morrow 2006; Lee and Van Zandt 2019; Murphy et al. 2009). Some of the disparities between renters and homeowners may be due to renters having less control over building risk mitigation (Murray 1999). Other reasons homeowners are more prepared include having better established communications channels created by living in their communities for longer, having more financial resources than renters, and being more likely to receive help from family (Lee and Van Zandt 2019). Indeed, homeowners are significantly more likely than renters to report having the resources they need to prepare (Mulilis, Duval, and Bovalino 2000).

Increasingly, attention is being paid to housing conditions (e.g., adequacy, year built, building type) and neighborhood characteristics (e.g., perceptions of crime) and their association with preparedness (Friedman et al. 2021; Lee and Van Zandt 2019). For example, households living in a building with two or more units, in housing that is moderately or severely inadequate, and in neighborhoods perceived to have serious crime are significantly less prepared than households living in single, detached housing units, in homes that are adequate, and in neighborhoods without serious crime, respectively (Friedman et al. 2021). Compared to mobile homes and multiunit buildings, residents living in single-family detached housing units are less likely to suffer damage from weather hazards (Hurlb et al. 2014). Disparities between neighborhoods impact household preparedness and disaster response and recovery (Van Zandt et al. 2012). For example, higher levels of neighborhood crime, especially in conjunction with other negative neighborhood qualities, can increase vulnerability to disasters by reducing collective efficacy (Sampson 2012). Reduced collective efficacy means less sharing of information and resources among neighbors (Hurlbert, Haines, and Beggs 2000; Klinenberg 2002).

Subsidized renters are more likely than unsubsidized renters and homeowners to live in older housing that is has worse housing conditions. According to Docter and Galvez (2019), 51% of public housing units were constructed before 1975, and some 28% of public housing properties were in some state of disrepair in 2019. However, controlling for differences in the age of the housing stock and variation in housing conditions may not be enough to account for preparedness disparities between those living in subsidized rental housing and those living in unsubsidized rental housing or owner-occupied housing.

The historical, structural processes underlying decisions about the location of subsidized housing in cities could play a role; 40% of public housing and 30% of LITHC units are built in areas that are vulnerable to natural hazards, like hurricanes, flooding, and tornadoes, compared to 27% of all rental units (PAHRC and NLIHC 2021). One study on Hurricane Harvey found that occupied subsidized housing units in Houston, Texas are disproportionately located in census tracts that experience flooding (Chakraborty et al. 2021). This disparity in exposure to hazards highlights the importance of considering spatial vulnerability.

In addition to being located near natural hazards, subsidized housing is in areas with little commercial development (Talen and Kocschinsky 2014). Between the 1950s and 1970s, public housing was intentionally located in poor, black neighborhoods, which have historically been avoided by commercial developers (Massey and Kanaiaupuni 1993). Housing voucher holders also reside in moderate-to-high poverty neighborhoods largely because of structural mechanisms like housing discrimination, insufficient information about rental opportunities, and inadequate supply of fair-market rate housing in nonpoor areas (Talen and Koschinsky 2014). Indeed, in some areas, housing voucher holders have even less access to amenities and services than public-housing residents because of the lack of public transportation (Talen and Koschinsky 2014).

The location of subsidized residents in areas that are more vulnerable to natural hazards and at the same time lack services and amenities will likely lead to subsidized residents being less prepared for disasters than unsubsidized renters and homeowners. Without access to adequate supermarkets and other retail establishments, it will be harder for subsidized residents, relative to unsubsidized renters and homeowners, to access the resources that they need to prepare, especially at affordable prices, like non-perishable food, water, and emergency supplies. Moreover, the lack of planning on the part of Public Housing Authorities (PHA) for disasters could further reduce the preparedness of subsidized housing residents. HUD’s own PHA Disaster Readiness and Preparation Guide (2016) recommends housing authorities actively encourage individual preparedness among residents as a part of their disaster preparedness planning (e.g., connecting residents to educational information; holding seminars and trainings). PHAs that overlook individual resident preparedness in their organizational planning may leave the most vulnerable residents, such as those with limited social networks and limited internet access or literacy (Li et al. 2023), devoid of information and time to prepare.

Hypotheses

The primary research question of this paper is: Do subsidized renters differ from unsubsidized renters and homeowners in their disaster preparedness? Based on the preceding discussion we can make several hypotheses. First, homeowners will be more prepared than renters regardless of their subsidy status. We expect this will remain true even when controlling for demographic, housing, and neighborhood characteristics as homeownership generally grants access to wealth that we cannot account for, and homeowners are likely to take every precaution possible to protect themselves and their homes in the event of a disaster because investment in their homes translates into wealth accrual². Second, subsidized renters will be less prepared than unsubsidized renters, at least in the descriptive analyses, because subsidized housing residents have characteristics which make them socially vulnerable and are associated with a lack of disaster preparedness. If subsidized renters continue to be less prepared than unsubsidized renters even after including controls for demographic and social characteristics in the multivariable analysis, the remaining difference could be attributable to the structural factors of the uneven spatial location of subsidized housing discussed just above.

Data and Methods

Data.

The American Housing Survey (AHS) provides information on the composition, size, and quality of the American housing stock. It is a biannual, nationally representative sample (U.S. Census Bureau 2020). This paper uses the 2017 AHS dataset because it includes questions about household disaster preparedness. The 2017 AHS Integrated National sample consists of 84,879 housing units (U.S. Census Bureau 2018). Importantly for the purpose of this paper, the AHS intentionally oversamples units receiving HUD rental assistance. While every household sampled was asked the set of core questions, which include household demographics, the rotating topical modules -- one of which is disaster preparedness -- used split sampling. This means only half of the households sampled were asked about their disaster preparedness. The split sample is created by randomly splitting the whole sample into two groups. The 2017 AHS split sample includes just over 34,000 housing units.

Measures.

Dependent variables.

The AHS asked a split sample of households a total of fourteen preparedness questions. The preparedness items included in this analysis from the disaster preparedness section of the AHS 2017 questionnaire are: 1) enough non-perishable food for 3 days; 2) at least 3 gallons or 24 bottles of water per person; 3) an emergency supply kit; 4) savings or credit card balances to meet evacuations expenses of up to $2,000; 5) access to a reliable vehicle for evacuation; 6) an electric generator (which is only asked of households in detached or multiunit buildings with four or less units); 7) access to vital financial information and contact numbers in an evacuation; 8) a communications plan; and 9) an established meeting point during an evacuation. The questions about having a communications plan and an established meeting point during an evacuation are only asked of households with two or more members.

Other sections of the survey asked respondents if they have property insurance and a carbon monoxide detector. Purchasing flood insurance is a preparedness variable used in some literature (Martins et al. 2019; Van Zandt et al. 2012), and flood insurance is among the disaster preparedness questions asked in the 2017 AHS. However, given the range of exposure to flooding between single family homes and high-rise apartments or condos, we propose that having any property or renters insurance, not just hazard-specific insurance, indicates a level of preparedness as it decreases social vulnerability to any type of risk (Martins et al. 2019). The Community Assessment for Public Health Emergency Response (CASPER) includes having a working carbon monoxide detector among their disaster preparedness variables (Murti et al. 2014). The AHS asks households if they have a carbon monoxide detector in the appliances section of the survey; we used it here as another indicator of preparedness.

The remaining 2017 AHS preparedness questions will not be included due to their lack of applicability to all households. The excluded questions relate to needing assistance for a pet and having a tornado safe room. While the AHS includes a number of key preparedness items, it does not ask households about having a battery-operated flashlight or radio, measures that have been used in other research (Kohn et al. 2012). Nevertheless, the breadth of the questions that we use here covers most of the aspects of preparedness used in other research (e.g., Friedman et al. 2021; Malmin 2021; Murphy et al. 2009).

Independent variables.

Our primary independent variable is housing tenure. It is composed of three groups: subsidized renters, unsubsidized renters, and homeowners. Our primary independent variable is a combination of the AHS variables HUDSUB, which distinguishes between HUD rental assistance recipients and non-assisted renters, and TENURE, which distinguishes between homeowners and renters. We create two dummy variables for the multivariable analysis, with subsidized renters serving as the reference group. Subsidized renters are defined as households: 1) living in public housing, which are units owned and operated by local public housing authorities; 2) receiving housing vouchers to cover all or part of their rent from the Housing Choice Voucher program; and 3) living in a multifamily unit that is considered to be project-based, subsidized housing whereby the landlord receives a government subsidy to provide affordable housing (HUD and U.S. Census Bureau 2017).³

Housing unit and neighborhood characteristics included as control variables are the year the unit was built; the type of housing structure; dwelling adequacy; perception of neighborhood crime; and perception of the neighborhood being at high risk for flooding and other disasters. The year a unit was built falls under three categories: before 1970, between 1970 and 1999, and 2000 or later. Housing type is organized into four categories, single-family detached, single-family attached, mobile home, and multiunit housing. Regarding adequacy, units are coded as adequate or inadequate, which combines units considered to be moderately or severely inadequate, based upon the condition and presence of plumbing, heating, electricity, wiring, and upkeep in the unit (HUD and U.S. Census Bureau 2017). Residents’ perceptions of serious crime being a problem in their neighborhood is used to gauge neighborhood crime. Agreeing that one’s neighborhood is at high risk for floods and other disasters is used as a measure of the neighborhood’s risk for disaster. Finally, the U.S. is divided into four regions, the Northeast, Midwest, South, and West, and we create dummy variables for these regions with the Northeast serving as the reference group in the multivariable analyses.

Householder and household variables included in the analysis are: householder: race and ethnicity, education, nativity status, marital status, gender, disability status, and age; and for the household: income, the presence of children and older adults, and the number of years in the unit. We restrict our analytical sample to householders of the following four racial and ethnic backgrounds because other categories have sample sizes that are too small: non-Hispanic Whites, non-Hispanic Blacks, Hispanics, and Asians. Household income is measured by an indicator variable that distinguishes households that have income at or above the poverty line or below the poverty line. Educational attainment is measured by four categories gauging the householder’s educational attainment – 1) less than a high school education, which is the reference group in our multivariable analyses; 2) high school diploma or GED; 3) some college; and 4) at least a college degree. Demographic factors are represented by three dummy variables indicating: 1) whether the householder is foreign born; 2) whether the household is headed by a married couple; and 3) whether children under 18 are present. We also use an indicator for the presence of older adults at/above 65 years old or younger than 65 years old. We include a dummy variable to measure disability status; if the household has at least one person who has difficulty hearing, seeing, dressing or bathing, doing errands, concentrating or remembering, or walking, the household is considered as having a disabled person. Another demographic measure we include is the householder’s age. Years spent living in the unit is also included, as the amount of time a household lives in their community may impact their preparedness (Lee and Van Zandt 2019).

Analytical Plan.

For several of our dependent variables and some independent variables, we have missing data. If we took a naïve approach like using a complete-case-only analysis, we would drop 7.6% of cases due to missing values. However, we employ multiple imputation (MI), consistent with the survey design, to avoid the adverse impact of a complete-case-only approach (see for example Little and Rubin 2019). Using SAS PROC MI (with the srmi option), we utilize the same approach discussed in detail in Friedman et al. (2021).

We conduct bivariate and multivariable analyses of these data. Bivariate analyses are used to examine how subsidized renters compare to unsubsidized renters and homeowners in terms of their disaster preparedness as well as their residential, socioeconomic, and demographic characteristics. Based upon the multiply imputed data, we use logistic regression analyses to examine whether any disadvantages in disaster preparedness persist for subsidized renters, relative to unsubsidized renters and homeowners, after controlling for the other variables.

Results

Bivariate analyses of each preparedness item by housing tenure are shown in Table 1. The percentages fulfilling each of the preparedness items are shown for the total sample, subsidized renters, unsubsidized renters, and homeowners in columns 1, 2, 3, and 5 respectively. Among all households (column 1), access to an evacuation vehicle is the most fulfilled resource-based item with 92.9% of the overall sample having one. The least fulfilled item is having a generator, with only 19.4% of households having one. Just over half of the total sample have an emergency preparedness kit (54.3%) and at least three gallons of water per person (59.7%). About 64% of households have a carbon monoxide detector. About three-quarters of households have property or renter’s insurance, and over three-quarters have evacuation funds of up to $2,000 (78.6%) and enough non-perishable food to last for at least three days (83.3%).

Table 1.

Housing Tenure Differences in Disaster Preparedness in the U.S., 2017

Disaster Preparedness Item	Percent of:
	Total	Subsidized Renters	Unsubsidized Renters	Sig.	Owners	Sig.
	(1)	(2)	(3)	(4)	(5)	(6)
Household has:
Resource based:
Enough non-perishable food for at least 3 days	83.3	80.5	77.2	*	86.3	***
At least 3 gallons of water per person	59.7	55.6	54.8		62.2	***
Prepared emergency supply kit	54.3	49.9	51.4		56.0	***
Evacuation funds of up to $2,000	78.6	29.2	64.7	***	87.8	***
Has property or renter’s insurance	74.6	17.1	40.6	***	93.7	***
Evacuation vehicle(s) available	92.9	63.7	87.4	***	97.1	***
Carbon monoxide detector	63.8	60.5	60.2		65.8	**
Generator1	19.4	3.6	10.3	***	22.5	***
Action Based:
Access to financial and contact information	83.2	74.1	80.3	***	85.0	***
Communications Plan2	26.8	32.8	27.4	*	26.4	**
Evacuation Meeting Location2	37.3	38.4	35.2		38.1
N	25956	1989	8082		15885

^***p<.001

^**p<.01

^*p<.05 -- ref. is unsubsidized renters; shading indicates difference between subsidized renters and owners are statistically significant at least at the .05 level

Notes: NH = Non Hispanic

¹This variable is only asked for households living in buildings with less than 5 units; N(total = 20637, subsidized renter = 876, renters = 4520, owners = 15241)

²This variable is only asked for households with two or more members;N(total = 18757, subsidized renter = 1011, renters = 5311, owners = 12435)

Columns 4 and 6 show whether the differences in the percentages between subsidized renters and a) unsubsidized renters and b) homeowners are statistically significant, respectively. Statistically significant differences between unsubsidized renters and homeowners are denoted by the shading of cells in column 3. With respect to the resource-based items, our analyses reveal that subsidized and unsubsidized renters are significantly less prepared on all items than homeowners consistent with previous research, and relative to unsubsidized renters, subsidized renters are significantly more likely to have enough non-perishable food to last for at least three days. However, subsidized renters are significantly less likely than unsubsidized renters to have evacuation funds of up to $2000, property or renter’s insurance, an evacuation vehicle, or a generator. Particularly large differences in preparedness are evident between subsidized renters and the other two housing tenure groups for the variables, evacuation funds, having property or renter’s insurance, and having a generator. Subsidized renters are half as likely as unsubsidized renters and are one-third as likely as homeowners to be prepared on those preparedness items.

Turning to the action-based items, for all households, column 1 of Table 1 shows that 83.2% of all households would have access to financial and contact information in the event of disaster, but fewer households have a communications plan (26.8%) and an evacuation meeting location (37.3%). Subsidized renters and unsubsidized renters are significantly less likely to have access to financial and contact information than homeowners. However, subsidized renters are significantly more prepared than unsubsidized renters and homeowners in terms of having a communications plan. With respect to having an evacuation meeting location, the only significant difference is between unsubsidized renters and homeowners, with the latter group being significantly more prepared than the former group (38.1% versus 35.2%). In summary, the results in Table 1 show that subsidized renters are particularly disadvantaged in their preparedness, relative to homeowners (with the exceptions of having a communications plan and evacuation meeting location). Subsidized renters are also significantly less prepared than unsubsidized renters but, on less items, than is the case relative to homeowners.

The differences in the preparedness of households based upon their housing tenure could be attributable to differences in their housing and residential characteristics as well as household and householder demographic and socioeconomic attributes. Table 2 examines the differences in these characteristics by housing tenure. Here we present the characteristics of the overall sample, subsidized renters, unsubsidized renters, and homeowners in columns 1, 2, 3, and 5, respectively. Columns 4 and 6 show whether the differences in the percentages between subsidized renters and a) unsubsidized renters and b) homeowners are statistically significant, respectively. Shaded cells in column 3 indicate that the difference between unsubsidized renters and homeowners is statistically significant.

Table 2.

Housing Tenure Differences in Housing, Household, and Householder Characteristics in the U.S., 2017

Characteristic	Percent of:
	Total	Subsidized Renters	Unsubsidized Renters	Sig.	Owners	Sig.
	(1)	(2)	(3)	(4)	(5)	(6)
Housing and Residential Characteristics
Year built
before 1970	40.0	47.6	43.2	*	38.1	***
1970–1999	41.1	39.6	38.4		42.4
2000 or later	18.9	12.8	18.4	***	19.5	***
Type of building in which unit is located
single-family home, detached	64.7	13.6	31.1	***	83.4	***
single-family home, attached	7.2	10.8	9.3		6.1	***
multi-family housing	22.5	75.0	55.2	***	4.2	***
mobile home, trailer, RV or other	5.5	0.6	4.4	***	6.3	***
Moderately or severely inadequate housing	4.8	8.0	7.6		3.3	***
Agrees neighborhood has serious crime	6.0	21.1	8.8	***	3.9	***
Neighborhood is at high risk for floods or other disasters	7.7	10.2	8.9		7.0	**
Region
Northeast	18.2	28.3	18.7	***	17.4	***
Midwest	22.5	21.6	18.9		24.3
South	37.5	33.6	36.9	*	38.0	**
West	21.8	16.6	25.5	***	20.3	**
Household and Householder Characteristics
Race/ethnicity
NH White	68.4	35.0	55.5	***	76.3	***
NH Black	13.1	42.9	18.1	***	9.1	***
Hispanic	13.8	19.6	20.8		10.2	***
Asian	4.7	2.4	5.6	***	4.4	***
Income at or below the poverty line	13.5	61.8	18.9	***	8.4	***
Householder education
less than high school degree	10.7	30.3	13.7	***	8.2	***
high school degree	24.8	32.8	26.9	***	23.3	***
some college	29.2	28.1	30.9		28.5
bachelor’s degree or more	35.3	8.7	28.5	***	40.0	***
Foreign-born	15.3	14.0	21.7	***	12.3
Household is a married-couple family	51.0	11.7	33.2	***	61.5	***
Household has children	30.4	35.0	32.2		29.3	***
Household has at least one person aged 65+	29.0	28.1	16.1	***	35.2	***
Householder is female	48.0	75.2	50.7	***	45.3	***
At least one person in the home is disabled	22.7	45.2	19.8	***	23.0	***
Householder age (mean)	52.2	52.7	44.3	***	56.0	***
Number of years household in the unit (mean)	12.3	6.9	4.7	***	16.2	***
N	25956	1989	8082		15885

^***p<.001

^**p<.01

^*p<.05 -- ref. is unsubsidized renters; shading indicates difference between subsidized renters and owners are statistically significant at least at the .05 level

Notes: NH = Non Hispanic

Source: Authors’ tabulations of the 2017 American Housing Survey

With respect to housing characteristics, subsidized and unsubsidized renters are significantly more likely to live in older housing stock and less likely to live in detached, single-family homes than homeowners. Just over 47% of subsidized renters and 43.2% of unsubsidized renters live in housing that was built before 1970, compared to 38.1% of homeowners. With respect to the type of housing, it is notable, although not surprising, that 75% of subsidized renters live in multi-family housing compared to 55.2% of unsubsidized renters and 4.2% of homeowners. In addition to that difference being significant between subsidized and unsubsidized renters, subsidized renters are significantly less likely than unsubsidized renters to live in detached single-family homes.

Table 2 shows that neighborhood quality and geographic location also vary by housing tenure. Subsidized and unsubsidized renters are significantly more likely than homeowners to report living in housing that is moderately or severely inadequate and in neighborhoods with serious crime and a higher risk of flooding. Both types of renters are also significantly more likely than homeowners to live in the Northeast. Subsidized renters are significantly more likely than unsubsidized renters to report that their neighborhoods have serious crime (2.4 times more likely) and live in the Northeast. Regarding geographic location, subsidized renters are significantly less likely to live in the West, relative to unsubsidized renters and homeowners.

The second half of Table 2 shows household and householder characteristics by housing tenure. There are notable differences in the racial and ethnic composition and socioeconomic status of the occupants by housing tenure. Subsidized and unsubsidized renters are significantly more likely than homeowners to be non-white, have less income, and lower education. Subsidized renters are significantly more likely than unsubsidized renters to be non-white, poorer, and less education. For example, non-Hispanic Blacks are significantly overrepresented in subsidized rental housing, composing 42.9% of the occupants in such housing, relative to only composing 18.1% of unsubsidized rental housing and 9.1% of owner-occupied housing. A significantly larger proportion of subsidized renters (61.8%) live at or below the poverty line compared to 18.9% of unsubsidized renters and 8.4% of owners. Almost one-third (30.3%) of subsidized renters do not have a high school diploma compared to 13.7% of unsubsidized renters and 8.2% of owners. These differences in racial and ethnic composition and socioeconomic status are likely to be associated with the differences in preparedness by housing tenure observed in Table 1.

There are other notable demographic differences shown in Table 2 that likely contribute to variation in preparedness by housing tenure. Subsidized and unsubsidized renters are significantly less likely than homeowners to live in married-couple families, have older adults, be older, and spend less years in their units. On the other hand, subsidized and unsubsidized renters are significantly more likely than homeowners to have children and be female. Relative to unsubsidized renters, subsidized renters are significantly less likely to be foreign-born⁴ and live in married-couple families. However, subsidized renters are significantly more likely than unsubsidized renters to have at least one person aged 65 and over, be female, have at least one person with a disability, be older, and live in the unit for more years. The magnitude of the differences between subsidized and unsubsidized renters on some of these characteristics is notable and could have implications for the differences in preparedness observed in Table 1. About one-third as many subsidized renters are married couples compared to unsubsidized renters, which could mean that subsidized renters have less resources than unsubsidized renters. At the same time, subsidized renters are 1.74 and 2.28 times more likely than unsubsidized renters to have at least one person in their households that are at least 65 years in age and has a disability, respectively.

What is the association between housing tenure and preparedness once we control for the variation in housing, residential, household, and householder characteristics? Table 3 reports the results, specifically the odds ratios, just for our key housing tenure variables from eleven separate logistic regression models of the resource- and action-based preparedness indicators, with all control variables included in the models. Appendix Tables 1 and 2 show the full results for the models of resource- and action-based outcomes, inclusive of control variables. Column 1 of Table 3 shows that relative to subsidized renters, unsubsidized renters are significantly more likely to fulfill four out of eight resource-based items, controlling for other factors. The odds of unsubsidized renters having evacuation funds of up to $2000, property or renter’s insurance, an evacuation vehicle and a generator are 2.18, 1.82, 1.54, and 1.87 times the odds, respectively, of unsubsidized renters. These results are consistent with those in the descriptive analyses shown in Table 1 and reveal that the control variables did little to attenuate these differences. Subsidized renters are significantly more likely than unsubsidized renters to have at least a three-day supply of nonperishable food (see column 1 of Table 3), consistent with the results in Table 1, and a carbon monoxide detector, a difference that emerges as significant in the multivariable analysis.

Table 3.

Selected Odds Ratios of Housing Tenure Variables from Separate Logistic Regression Models of Disaster Preparedness Outcomes in the U.S., 2017

Dependent variables1	Key Independent Variables (Subsidized Renters ref.)
	Unsubsidized Renters	Sig.	Owners	Sig.
	(1)		(2)
Resource based:
Enough non-perishable food for at least 3 days	0.783	**	1.014
	(0.650, 0.943)		(0.823, 1.249)
At least 3 gallons of water per person	1.128		1.433	***
	(0.970, 1.312)		(1.214, 1.692)
Prepared emergency supply kit	1.041		1.140
	(0.896, 1.210)		(0.967, 1.344)
Evacuation funds of up to $2,000	2.183	***	5.423	***
	(1.820, 2.618)		(4.437, 6.628)
Has property or renter’s insurance	1.815	***	40.121	***
	(1.494, 2.204)		(31.944, 50.391)
Evacuation vehicle(s) available	1.537	***	3.145	***
	(1.275, 1.854)		(2.463, 4.015)
Carbon monoxide detector	0.769	**	0.985
	(0.657, 0.900)		(0.828, 1.172)
Generator2	1.866	**	2.672	***
	(1.107, 3.146)		(1.587, 4.498)
Action Based:
Access to financial and contact information	1.057		1.304	**
	(0.886, 1.261)		(1.069, 1.592)
Communications Plan3	0.915		0.916
	(0.739, 1.134)		(0.728, 1.153)
Evacuation Meeting Location2	0.897		0.912
	(0.730, 1.102)		(0.732, 1.137)

^***p<.001

^**p<.01

^*p<.05

Source: Authors’ tabulations of the 2017 American Housing Survey

¹Each row is a separate model for the dependent variable shown in that row.

²This variable is only asked for households living in buildings with less than 5 units; N(total = 20637, subsidized renter = 876, renters = 4520, owners = 15241)

³This variable is only asked for households with two or more members; N(total = 18757, subsidized renter = 1011, renters = 5311, owners = 12435)

Relative to homeowners, how do subsidized and unsubsidized renters compare in terms of their preparedness on resource-based outcomes in the multivariable analysis? Column 2 of Table 3 reveals that several of the significant differences remain, compared to Table 1, with subsidized renters being disadvantaged in their preparedness in terms of having enough water, evacuation funds, property or renter’s insurance, an evacuation vehicle, and a generator, compared to homeowners. The differences between subsidized renters and homeowners are no longer significant in terms of having an adequate supply of non-perishable food, an emergency supply kit, and a carbon monoxide detector. Appendix Table 3 shows that unsubsidized renters are significantly less likely to be prepared on all resource-based items than homeowners, thereby revealing that control variables had little impact on attenuating these differences from the bivariate analyses conducted in Table 1. These findings align results from previous research that show that homeowners are more prepared than all renters (Friedman et al. 2021; Murphy et al. 2009; Rivera 2020).

In terms of action-based preparedness items, column 1 of Table 3 reveals no statistically significant differences between unsubsidized and subsidized renters, in contrast to the descriptive analyses in Table 1 that showed two significant differences. Conversely, column 2 of Table 3 shows that the odds of homeowners having access to financial and contact information are 1.3 times more likely than for subsidized renters, controlling for relevant factors. This significant difference therefore persisted in the multivariable analysis. A similar significant difference is evident between homeowners and unsubsidized renters (see column 2 of Appendix Table 3). The difference between owners and subsidized renters in terms of having a communications plan found in Table 1 is no longer statistically significant in the multivariable analysis in column 2 of Table 3. Appendix Table 3 shows a similar set of findings regarding having an evacuation meeting location; the difference between homeowners and unsubsidized renters on this variable is no longer significant in the multivariable analysis.

Taken together, the results in Table 3 and Appendix Table 3 reveal that in terms of resource-based outcomes, subsidized renters are generally the least prepared relative to the other two housing-tenure groups. Owners are generally the most prepared housing tenure group, and unsubsidized renters fall in between subsidized renters and homeowners in terms of their resource-based preparedness. However, in terms of action-based preparedness items, there are fewer differences by housing tenure.

Turning to the associations between our control variables and preparedness outcomes, we provide a brief discussion of results pertaining to our resource- and action-based preparedness outcomes among our control variables. Regarding housing and residential characteristics, residents living in housing built since 1970 are generally more prepared in terms of resources than those living in older housing. In terms of action-based outcomes, there are less consistent differences by the year the housing was built. Households living in single-family units (i.e., the reference group in Appendix Tables 1 and 2) are more likely to be prepared in terms of resource- and action-based preparedness items than households living in other unit types. Those living in moderately or severely inadequate housing as well as those living in neighborhoods with serious crime are generally significantly less likely to be prepared in terms of resource- and action-based items than those living in adequate housing and in neighborhoods without serious crime, respectively.

Interestingly, households that consider their neighborhood to be at risk for disaster are not necessarily more or less likely to prepare for disasters than those that do not consider their neighborhood to be at risk. Only on two out of eight resource-based items and two out of three action-based items are those who consider their neighborhoods at risk for a disaster significantly more likely to prepare than those who do not consider their neighborhoods at risk for a disaster. The lack of significant associations between perceived neighborhood risk for disaster and some emergency preparedness items is consistent with studies that show that households may fail to prepare because they feel a lack of control over these severe climate hazards (Han et al. 2017). The findings are also consistent with research that shows that individuals’ actual experiences with previous severe weather and climate disasters are more influential in their subsequent preparation for disasters than their perception of whether their neighborhood is at risk for a disaster (Han et al. 2017; Hernández et al. 2018; Tomio, Sato, and Mizumura 2012). Resource- and action-based preparedness also varies by region, but there is little consistency in these patterns by region.

How does preparedness vary by the demographic and socioeconomic characteristics of householders and households? Appendix Tables 1 and 2 show that non-whites are generally disadvantaged, relative to Whites, in fulfilling their resource- and action-based preparedness outcomes, controlling for other relevant factors. Households living at or below the poverty line have fewer financial resources to prepare for disasters. Unsurprisingly, such households are significantly less likely to have evacuation funds, insurance, an evacuation vehicle, a carbon monoxide detector, and a generator. Yet, households at or below the poverty line are significantly more likely to have water stockpiled (see column 2 of Appendix Table 1) as well as action-based preparedness of having a communication plan for households with two or members (see Appendix Table 2, column 2).⁵ In general, education is positively associated with both types of preparedness, which is consistent with previous literature (Al-rousan et al. 2014; Kohn et al. 2012).

Other demographic factors, besides race and ethnicity, are also associated with resource- and action-based preparedness. In general, foreign-born individuals, are less prepared than native-born individuals. Married households are significantly more likely to fulfill almost all resource-based and action-based preparations. Households with a disabled person are less likely than those without a disabled person to fulfill numerous items (columns 2–7 of Appendix Table 1 and column 1 of Appendix Table 2) when controlling for other factors. These findings for disability status replicate those in previous research (e.g., Zamboni and Martin 2020). The disaster preparedness literature generally finds that older adults are more prepared for disasters (Ablah, Konda, and Kelley 2009; Zamboni and Martin 2020). Our results are generally consistent with previous research in that older households are significantly more likely to be prepared in terms of resource-based outcomes than younger households (see the results for householder age in Appendix Table 1); for action-based preparedness, however, we do not observe significant differences by householder age (see the results for householder age in Appendix Table 2).

Discussion and Conclusion

Subsidized housing residents often fall into many co-existing socially vulnerable categories that place them at higher risk of negative disaster outcomes than unsubsidized renters and homeowners. The primary purpose of this paper was to establish whether residence in subsidized housing is associated with greater levels of disaster preparedness of households relative to those living in unsubsidized rental housing and owner-occupied housing. The results indicate that subsidized renters are less prepared in terms of resource-based items than unsubsidized renters, controlling for other factors including the socially vulnerable statuses that prevent them from being more prepared. Subsidized renters were less likely to have evacuation funds, insurance, access to an evacuation vehicle, and a generator. The only items subsidized renters were significantly more likely than unsubsidized renters to fulfill were a three-days of nonperishable food and having a carbon monoxide detector. With respect to owners, our results echo those of previous research that find that owners are more prepared than both groups of renters, controlling for other factors (Murphy et al. 2009; Rivera 2020). Relative to subsidized renters, our multivariable analyses reveal that homeowners are significantly more prepared on several resource- and action-based items except for having enough non-perishable food for at least three days, having an emergency supply kit, a carbon monoxide detector, a communications plan, and an evacuation meeting location. On those items, there are no significant differences.

Our results support for all our hypotheses. Homeowners were found to be significantly more prepared for disasters on most outcomes than unsubsidized renters, controlling for other characteristics. We found support for the hypothesis that the lack of preparedness among subsidized renters, relative to unsubsidized renters, is attributable to their social vulnerability. The multivariable results in Table 3 revealed that difference between subsidized and unsubsidized renters in terms of having access to financial and contact information is no longer statistically significant after controlling for the differences in housing, residential, socioeconomic, and demographic characteristics. The results in Table 3 also show that some of the differences in preparedness between owners and subsidized renters become insignificant after controlling for factors indicating differences in social vulnerability between the two groups. We also find support for our third hypothesis. Even after controlling for the fact that households in subsidized rental units are the most socially vulnerable, they are still significantly less prepared than unsubsidized renters and homeowners in terms of evacuation funds, renter’s insurance, the availability of an evacuation vehicle, and having a generator. Thus, subsidized renters’ lack of preparedness is not simply due to their socially vulnerable statuses but is likely also due to structural factors surrounding the operation and location of subsidized housing units. In addition, relative to unsubsidized renters, the significantly higher preparation levels of subsidized renters in terms of their three-day food supply and having a carbon monoxide detector persist in the multivariable analyses. These results also suggest a potentially positive impact that structural factors may play in contributing to their preparedness on these items.

While HUD provides PHAs with a Disaster Readiness and Preparation Guide as well as an online Disaster Planning, Response, and Recovery Toolkit, based upon our results, it appears that the effective implementation of these resources by PHAs at the household level is mixed (HUD 2016; HUD 2021a). Though both planning tools focus heavily on disaster preparation for PHAs as organizations (i.e., maintaining staff contact lists, developing continuity plans, establishing community partnerships), the Disaster Readiness and Preparation Guide includes a section on what PHAs can do to prepare their residents. Suggested preparedness actions include conducting drills, offering first aid classes to residents, informing residents of the location of shelters, and posting the location of emergency radios and other emergency supplies. The Disaster Planning, Response, and Recovery Toolkit includes disaster preparedness checklists for PHAs to provide to residents. These checklists include: evaluating financial preparedness for a disaster; creating a household emergency plan; stockpiling emergency supplies; and providing information on how to prepare for specific types of disasters.

More scrutiny of PHA disaster preparedness measures may be needed to ensure that best practices are being followed. It is also important to consider how policies and funding impact the ability for PHAs to prepare their residents. For example, PHAs are provided with funds for resident participation activities for the purpose of improving residential quality of life and to promote a positive living environment (HUD 2022b). Yet, there is no specific funding earmarked for PHAs to use on disaster training and resources for residents. Without such funding, it is likely very difficult for PHAs implement the planning tools recommended in the Disaster Readiness and Preparation Guide and the Disaster Planning, Response, and Recovery Toolkit. FEMA offers Preparedness Grants to develop and maintain state, local, tribal, and territorial capabilities to prevent and recover from disasters and emergencies (Homeland Security and FEMA 2022). HUD could offer PHAs similar grants to improve disaster planning among their residents. Such a grant program may also allow HUD to track disaster preparedness and mitigation efforts among PHAs.

Also, it may be too early to track the effectiveness of these policies. After Hurricane Sandy, the Office of Housing Counseling (OHC) created the Disaster Assistance Response Team (DART) to assist and support housing counseling agencies in disaster areas (HUD 2017). Since 2018, the OHC has created and provides guidance and training on how agencies can educate clients on emergency and disaster preparedness as well as created the Housing Counseling Disaster Recovery Toolkit website (HUD 2017; HUD 2021b). Because our data are from 2017, we may not be able to see the full effectiveness of these programs.

Notably, HUD is making strides to improve access for subsidized renters to carbon monoxide detectors. Notice 2022–01 announced that all federally assisted units are required to have carbon monoxide detectors by the end of 2022 (Fitzhugh 2022). Our results, which are based upon 2017 data, find that subsidized housing units are significantly more likely than unsubsidized renters to have carbon monoxide detectors, and there is no difference between subsidized renters and owners in having these detectors, controlling for other factors. These results suggest that there were likely earlier steps taken by HUD to install carbon monoxide detectors and that with the policy in effect at present, the numbers should continue to increase.

In addition to showing the importance of subsidized housing in being associated with disaster preparedness, our results revealed the significance of housing and residential characteristics, corroborating the results of recent research (Friedman et al. 2021; Lee and Van Zandt 2019). The year the housing is built, the type of unit or structure of the housing, and the perceived social disorder in the neighborhood are all significant correlates of preparedness. These results suggest that the distribution of residential quality likely parallels the distribution of vulnerability to disasters and the ability of residents to adequately prepare.

Our study is not without limitations, and clearly more needs to be done to identify why subsidized renters are less prepared than unsubsidized renters and owners. The 2017 AHS data are limited in evaluating the direct impact of PHA policies on the disaster preparedness of residents because for PHAs, the 2017 AHS questionnaire does not contain information as to whether they follow the recommendations provided to them by HUD (e.g., taking a CPR class or asking if household with older or disabled adults are signed up on a needs registry). It remains unclear if PHA-level data on disaster planning and preparedness programming for residents is aggregated by HUD. If such data exist, future research could merge it with restricted-individual AHS data to compare disaster preparedness among residents from different PHAs. Another limitation of the data is that the HUD Disaster Readiness and Preparation Guide was published in 2016, and the AHS data used in this paper were collected in 2017. PHAs may not have had sufficient time to implement their disaster preparedness programming before AHS responses were collected. Future research should therefore use more recent data. Furthermore, the Housing Counseling Disaster Recovery Toolkit was not yet developed at the time of the 2017 AHS so its impact on residential preparedness is not measured by this study (HUD 2021b). Additionally, causality cannot be drawn from the cross-sectional data or statistical methods used in this paper. Finally, we do not know the exact geographic location of the residents in the public use file of the AHS thereby limiting our ability to merge data on actual disaster exposure to the AHS individual-level data gauging preparedness.

Natural disasters exacerbate an already overburdened supply of various types of HUD subsidized rental housing as well as place an already vulnerable population at increased risk for negative outcomes (Davlasheridze and Miao 2021). Future research would profit from linking more information about PHAs in terms of their disaster readiness to individual-level data from the AHS. Moreover, research should be done using the restricted access version of the AHS data so that geographic location can be used, and information about the exposure to disasters can be incorporated for all households. Future research should also compare preparedness across different types of subsidized housing (e.g., Public Housing and HCV). Finally, given that subsidized renters were found to be less prepared for disasters, an audit of PHA disaster management plans that focuses on how PHAs assist households with their preparedness could be conducted to evaluate how disaster preparedness among subsidized renters can be improved.

The impacts of climate change are devastating, particularly as evidenced by the recent damage caused by Hurricane Ian. In this era of increasing billion-dollar disasters, the results in this study make clear that subsidized renters are at a significant disadvantage with respect to their preparedness for such severe climate events. This research serves as a starting point for future studies on this topic. Preparation – in the form of financial resources, access to generators and vehicles, and in terms of having action plans – is critical for reducing hospitalizations and the loss of lives due to power outages and the impacts of major storms, flooding, and wildfires. Housing policy makers and scholars should continue to build evidence to best prepare subsidized renter households for the next disaster.