Rental Assistance and Adult Self-Rated Health

Rental assistance, in the form of vouchers and project-based subsidized housing, is a primary source of affordable housing for low-income Americans, given a growing and severe shortage of private-market rental units. However, due to supply constraints, fewer than one in four eligible households receive this kind of assistance. In this paper, we examine associations between receipt of rental assistance and self-rated health among a sample of 400 low-income adults in one U.S. city. We find that individuals who currently receive rental assistance have lower odds of reporting poor or fair self-rated health than individuals who are currently on rental assistance waiting lists. These relationships persist after adjusting for factors that affect access to rental assistance and are not significantly modified by criminal justice history. Our findings suggest that the current unmet need for rental assistance may contribute to poor health among low-income Americans.

Rental assistance, provided by the U.S. Department of Housing and Urban Development (HUD) in the form of rental vouchers and project-based housing, is a primary source of affordable housing for low-income U.S. households given a growing shortage of affordable private-market rental units.¹ However, due to supply constraints, fewer than one in four eligible households receive this kind of assistance.² An emerging body of research suggests that this unmet need for rental assistance may have negative health consequences for the millions of Americans who are currently on waiting lists for assistance or are unable to access these waiting lists.^3–6

Housing is a well-established determinant of physical, mental, and behavioral health, ^7–9 and rental assistance is likely to improve housing access.¹⁰ First, by setting rent at 30% of an individual’s income, rental assistance can make housing affordable, freeing up financial resources for other health investments.^11,12 By making housing affordable, rental assistance can also prevent eviction and other forced moves that are associated with negative health consequences^13,14 and can support residential stability that is health promoting.¹⁵ Additionally, some research suggests that rental assistance can provide access to better housing quality than its recipients would otherwise be able to afford.^5,16 Finally, by facilitating access to stable and affordable housing, rental assistance can prevent homelessness and its well-documented negative health sequelae.⁷

While a large literature documents the relationship between housing and health, few studies have examined the health effects of rental assistance specifically. The majority of studies that do address rental assistance have focused on child well-being, providing some evidence that rental assistance contributes to reduced food insecurity, improved nutritional status, and improvements in behavioral and economic outcomes for children, though findings are not consistent across studies or sub-populations.^17–20 Research on rental assistance and adult health is far more limited. A few recent studies have begun to address this gap by combining multi-year HUD data with nationally representative health data from the National Health Interview Survey (NHIS). This data linkage allows researchers to compare individuals who receive rental assistance with a group of individuals who are unassisted, but according to HUD records, go on to obtain rental assistance within 2 years of when the health data are collected. These analyses find that public housing residents report better self-rated health and less psychological distress than those who move into public housing within 2 years of the NHIS survey.³ Other analyses find that recipients of all forms of rental assistance are are less likely to report an unmet need for health care due to cost than unassisted adults who go on to receive assistance within 2 years.²¹

These recent studies suggest health costs associated with unmet need for rental assistance. However, more research is needed to understand how these effects vary across groups and populations. Furthermore, these prior studies exclude individuals who may need rental assistance but never receive it, either because they never apply for assistance in the first place, or because they are deemed ineligible when they reach the top of the waitlist. This latter group may also experience an unmet need for rental assistance and related health consequences.

One group that may face challenges in accessing rental assistance are individuals recently released from prison. Though restrictions vary considerably across housing authorities, prior incarcerations, convictions, and even arrests can affect eligibility for rental assistance.²² In most cases, a criminal record or incarceration history does not constitute an automatic ban from rent-assisted housing.²³ However, policies that make admission decisions with regards to criminal justice history on a case-by-case basis are likely to serve as barriers to housing for criminal justice involved individuals who may face lengthy and labor intensive appeal processes, the outcomes of which may be influenced by criminal justice stigma.^24,25 Furthermore, prior qualitative work suggests that even when individuals are technically eligible for rental assistance, they may perceive their criminal records as disqualifiying factors.^25,26 Individuals who are recently released from prison face many barriers to securing private market rental housing including cost and background checks. ^27,28 Given these barriers, rental assistance may be a particularly valuable resource and particularly relevant to the health of those who are recently released from prison.

To examine the relationship between rental assistance and health among those with and without recent incarceration histories, we draw on data from a cohort study of 400 low-income adults, which was designed to examine the intersections of housing, mass incarceration, and health. We test the hypothesis that individuals on waiting lists for rental assistance will be more likely to report poor or fair self-rated health than individuals who receive rental assistance, adjusting for factors that affect admission into rent-assisted housing. We also examine whether associations between rental assistance receipt and self-rated health are modified by incarceration history.

Methods

Study Design and Setting

Our analyses draw on data from the Justice Housing and Health Study (JustHouHS), which is set in New Haven, CT, a city with approximately 130,000 residents. Like many US cities, New Haven is experiencing an affordable housing shortage. In 2016, 55% of New Haven renters spent more than 30% of their income on rent and 80% of New Haven renters in the lowest income quintile were severely cost-burdened, spending more than 50% of their income on rent.²⁹

Given these high rents, rental assistance is an important component of New Haven’s affordable housing landscape. In 2016, 9,153 New Haven households and 19,221 individuals received HUD funded rental assistance in the form of Housing Choice Vouchers, traditional public housing, project-based Section 8.³⁰ In addition to HUD funded assistance, in 2016, 692 New Haven households received state funded rental assistance through the Rent Assistance Program (RAP) administered by the Connecticut Department of Housing. A small number of New Haven residents also receive long-term rental assistance through supportive housing programs, which are often funded by HUD, and are designated for individuals who are chronically homeless, living with HIV/AIDS, recovering from addiction, or who have mental illness.

JustHouHS is a survey of low-income New Haven residents designed to examine the intersection of housing, mass incarceration, and health. All data collection and recruitment procedures were approved by the Yale Institutional Review Board. JustHouHS participants were recruited using a combination of flyers posted throughout the New Haven community (e.g., bus stops, clinics, public libraries, social service organizations), outreach from service providers, and snow-ball sampling. To be eligible for the study, all participants had to meet the following criteria: be over 18 years of age, a resident of the city of New Haven, and have no household members already enrolled in the study. Additionally, in order to obtain a low-income sample, eligibility was restricted to individuals who had either:1) received housing or food assistance in the past year 2) were Medicaid recipients 3) were homeless or 4) resided in low-income census tracts (tracts where more than 20% of residents lived below the federal poverty level). Given the study’s interest in the intersection of mass incarceration and health, the sample was stratified to include 200 individuals released from prison or jail in the last year and 200 individuals who were not recently released from prison or jail but may have had prior incarcerations. Incarceration history was verified using data from the Connecticut Department of Corrections. Interested participants (N=616) contacted the study office and completed an eligibility screening over the phone or in person. To reach a sample size of 200 for each arm (recent incarceration and no recent incarceration), eligible participants were enrolled until their arm of the study was full.

Data Collection

Participants completed Qualtrics surveys on computers in the study office. The surveys took between 1 and 2 hours to complete and participants were compensated with a $50 gift card. The analyses presented here rely on baseline survey data collected between October 2017 and March 2018.

Measures

Our primary independent variables are measures of current rental assistance status. We create three mutually exclusive categories: those receiving assistance, those on a waiting list for assistance and not currently receiving another form of assistance, and those who are neither on a waiting list nor receiving assistance. Previous work suggests that self-report of rental assistance can be unreliable,³¹ in part because of the multiple and inconsistent terms individuals use to denote participation in assistance programs. We improve on prior surveys by asking participants if they have ever applied for, are currently receiving, or are on a waitlist for each form of rental assistance that is available in New Haven.

Our independent variable is self-rated health, assessed using the standard question, “Would you say your health in general is: excellent, very good, good, fair, or poor.” Self-rated health is shown to be a reliable measure of global physical and mental health and is highly predictive of morbidity and mortality.³² We dichotomize responses to examine the odds of reporting “poor or fair” self-rated health relative to all other categories. This dichotomization is consistent with the approach used in many recent studies that have examined housing and other social variables as predictors of self-rated health.^3,33

We include in our analyses factors that may provide preferential access to rental assistance receipt and are also likely related to health. In particular, a documented disability may provide preferential access to rental assistance given that some housing is set aside for individuals who have disabilities.³⁴ Receipt of disability benefits is also likely associated with poor self-rated health. Thus, disability may act as a negative confounder, attenuating the relationship between rental assistance and health. We adjust for disability status using a sequence of questions that first asks whether the individual has “ever applied for disability from the Social Security Administration” and then asks whether this application has ever been approved. We also include a measure of age as a continuous variable given that age is likely to be associated with worse self-rated health and greater access to rental assistance. Older adults may have access to senior public housing that is unavailable to younger adults. Additionally, older adults may have received assistance in an era where this resource was more plentiful. We also include a measure of whether participants live with children under the age of 18 given that children can provide preferential access to rental assistance.³⁵

We also include factors that may serve as barriers to rental assistance receipt. In particular, having a criminal justice history may create real or perceived barriers to obtaining rental assistance.^22,25 To account for this, we include two measures. First, we include a question that asks whether the individual was released from prison in the last 2 years. We choose this two-year time frame, rather than the one-year time frame used in sampling for JustHouHS, given evidence that incarceration can affect access to rental assistance beyond the period of one-year post-release.²⁵ We also include a measure of whether the individual has ever been convicted of a felony. Finally, we account for reported drug use in the last 30 days which could create a barrier to rental assistance receipt.

We include demographic variables that may be associated with both health and rental assistance. We assess race ethnicity using a two-part question that asks about participants’ racial identity and whether they identify as Hispanic or Latino. We assess employment through a question that asks whether and how much the participant has worked in the last 6 months. We report employment as a dichotomous variable of any versus no employment in the last 6 months. Additionally, we assess total monthly household income by combining all of a participant’s reported sources of income (e.g., employment, cash benefits) with those of other household members. We also calculate whether participants’ household incomes fall below HUD’s definition of low-income (80% of the area’s median income (AMI) for a given household size).

Analyses

In our analyses, we first compare characteristics of rent-assisted, waitlisted, and the neither rent-assisted nor waitlisted group (hereafter referred to as “neither”). We use ANOVA to test for statistically significant differences across all groups. Next, we estimate logistic regression models that predict the odds of reporting either poor/fair self-rated health. We first examine self-rated health as a function of rental assistance and waitlist status (model 1). We then add basic demographic factors (model 2). Next, we include additional factors that may affect access to rental assistance (model 3).

Additionally, we run a series of sensitivity analyses to test the robustness of our findings. First, we run analyses excluding 16 individuals who report that the type of rental assistance they receive is some form of supportive housing. These individuals may be receiving different benefits than individuals in traditional HUD programs. They also may have unique health needs. Second, we run analyses excluding 49 individuals who report living with someone who has rental assistance. Six of these individuals are counted as rent assisted in our sample (they answered affirmatively to receiving rental assistance and also answered affirmatively to living with someone who receives rental assistance). While living with someone who has rental assistance may be beneficial to health, the presence of these individuals in the sample complicates our ability to examine the effect of actually receiving this assistance. Finally, it is possible that any health advantage associated with rental assistance is related to the fact that many unassisted participants in our sample are homeless. While preventing homelessness is one mechanism through which rental assistance may improve health, we also want to consider whether there is a health benefit of rental assistance even among those who are housed. To do so, we run additional analyses that exclude the 72 individuals in our sample who report that their current place of residence is “homeless” or “homeless shelter.” Finally, in an additional set of models, we examine interactions between rental assistance and criminal justice variables. These interaction models include only covariates that were significant in model 3 above due to concerns about statistical power.

Results

Table 1 describes our sample and examines differences between rent-assisted, waitlisted, and the neither group. Seventy-three participants received rental assistance. This included 27 recipients of Housing Choice Vouchers, 12 living in project-based housing reserved for senior or disabled adults, three in traditional public housing, seven recipients of state rental assistance vouchers (RAP), five receiving assistance designated for persons living with HIV/AIDS, sixteen living in supportive housing, and three in other programs.

Table 1.

Descriptive Statistics

	Receiving Assistance	Waitlisted (and unassisted)	Neither waitlisted or assisted	Total % (N)	P
N	73	100	227	400
% Poor SRH	20.5	31.0	22.5	24.25	.181
Demographics
Mean Age	50.0	46.2	43.3	45.2	.000
%Male	40.0	61.0	79.7	67.8	.000
% NH Black	64.4	61.0	52.4	56.8	.123
% NH White	15.1	15.0	26.9	21.8	.017
% Hispanic/Latino	15.1	16.0	15.9	15.8	.98
% Other	5.5	8.0	4.8	5.8	.527
Mean household income*	$859.19	$759	$753	$775	.88
% Employed	42.5	53.0	49.8	16.5	.384
Potential barriers and facilitators to rental assistance
% Ever received disability	41.1	29.0	14.5	23.0	.000
% With children in home	17.8	14.0	11.0	13 .0	.307
Ever Felony	45.2	63.0	69.6	63.5	.001
Recent Incarceration	19.2	56.0	61.7	52.5	.000
Drugs 30 days	20.5	33.0	25.6	26.5	.166

^*Due to missing data N=346

The average age of our sample is 45. The groups differ significantly by age, with older ages in the rent-assisted group, likely reflecting access to senior public housing developments and cohort differences in rental assistance access. Our sample is more than two thirds male, likely reflecting JustHouHS’s over-sampling of individuals who were recently released from prison. There are also significant gender differences between the groups, with men concentrated in the neither group and women concentrated in the assisted group. The racial composition is similar across the three groups, with the exception that White participants are concentrated in the “neither” group. The average household monthly income was low ($775) and only half of the sample had worked in the last 6 months. Additionally, 98% of reported household incomes fell below HUD’s definition of low-income (80% of the AMI for their household size). There were no significant differences between the three groups with respect to employment and household income. Twenty-three percent of the sample had an approved application for disability benefits, and as expected, disability receipt was significantly more common among those receiving rental assistance. Only 13% of the sample had children under 18 living in their household and there were not significant differences across groups.

Given the deliberate over-sampling of recently released individuals, criminal justice involvement was common with 52% of the sample reporting an incarceration within the last two years and 63.5% reporting a prior felony conviction. As expected, criminal justice involvement varied significantly across the three groups, with the highest percentages in the neither group and the lowest in the rent-assisted group. One quarter of the sample had used any drugs within the past 30 days and there were not significant differences across the three groups.

Table 2 includes results from logistic regression models that report the odds of poor or fair self-rated health as a function of rental assistance and covariates. In the unadjusted model (model 1) there were not significant differences between rent assisted groups, though odds ratios suggest increased odds of poor or fair self-rated health associated with being on a waiting list. In model 2, we add basic demographics from Table 1, with the exception of income which we exclude due to missing data and because there were no significant differences across groups. This adjusted model indicates a significant benefit of rental assistance relative to the waitlist group. When other factors than can affect admission to rent-assisted housing are added in model 3, the effect size increases such that the odds of poor or fair self-rated health are 3.03 greater for the waitlist group relative to the rent-assisted group. In this model, as expected, disability is significantly associated with increased odds of reporting poor or fair health. Recent incarcerations and felony convictions are not significantly associated with self-rated health in this adjusted model.

Table 2.

Odds of Poor or Fair Self-Rated Health

	Model 1		Model 2		Model 3
	OR	95% CI	OR	95% CI	OR	95% CI
Rental Assistance
Waitlist	1.74	(0.85–3.53)	2.40 **	(1.14–5.07)	3.03***	(1.38– 6.64)
Neither	1.12	(0.59–2.14)	1.74	(0.85–3.54)	2.35**	(1.11–4.99)
Assisted	Ref	Ref	Ref	Ref	Ref
Demographics
Gender
Male			0.48***	(0.29–0.82)	0.60*	(0.33–1.05)
Female			Ref		Ref
Age (years)			1.04**	(1.02–1.06)	1.03*	(1.00–1.05)
Race ethnicity
NH Black			0.48***	(0.27–0.87)	0.44**	(0.24–0.87)
Hispanic			0.70	(0.33–1.49)	0.72	(0.34–1.55)
Other			0.41	(0.12–1.38)	0.43	(0.12–1.45)
NH White			Ref	Ref	Ref
Employment
Yes			.869	(0.54–1.14)	1.09	(0.65–1.83)
No			Ref		Ref
Potential barriers and facilitators of rental assistance (ref=no)
Disability					2.41***	(1.32–4.49)
Child in home					0.53	(0.22–1.25)
Recent Incar.					0.64	(0.34–1.19)
Felony					0.72	(0.39–1.35)
Recent Drugs					1.05	(0.60–1.90)
N	400		400		400

^***p<.01

^**= p <.05

^*P <.01

Table 3 provides results from sensitivity analyses where the sample is restricted to exclude individuals living in supportive housing, or individuals who are currently homeless, or individuals who are living with someone who receives rental assistance. In all models, the associations between rental assistance and health remain qualitatively similar to those in Table 2, though when homeless individuals are excluded the odds ratio for the neither group is no longer statistically significant. The interactions for criminal justice variables (results not shown) were not statistically significant.

Table 3.

Sensitivity Analyses^a

	Excluding participants living in supportive housing		Excluding participants living with someone in rent-assisted housing		Excluding currently homeless participants
	N=384		N=351		N=328
Waitlisted	2.86**	(1.31–6.91)	3.05***	(1.35–6.87)	3.32***	(1.45–7.59)
Neither	2.22*	(1.05–5.28)	2.34**	(1.07–5.11)	1.85	(0.86–4.00)
Assisted	Ref		Ref		Ref

^***p<.01

^**= p <.05

^*P <.01

^aModels adjusted for all covariates included in Table 2, Model

Discussion

In our analyses of data from 400 New Haven residents, we find that individuals who currently receive rental assistance have lower odds of reporting poor or fair self-rated health than individuals who are currently on waiting lists for rental assistance. These relationships persist, and are in fact strengthened, after adjusting for factors that can affect access to rental assistance including: age, disability, having children in the household, a recent incarceration, a prior felony conviction, and recent drug use. Our findings add to recent research that finds positive associations between the receipt of rental assistance and health ^3,17 and contribute to a growing body of evidence regarding the potential health costs of unmet need for this resource. ^4,16,36 Currently, fewer than one in four eligible households receives rental assistance.² Waiting lists average over 2 years nationally, but in many locations, are longer or closed to new applicants.²

The observed relationship between receipt of rental assistance and better health may reflect improved access to housing that assistance provides in an otherwise unaffordable rental market. Recent data from the Urban Institute indicate that in New Haven County, without rental assistance, there would be zero affordable and available housing units for individuals earning less than 30% of the AMI, suggesting few unsubsidized housing options for participants in our sample.³⁷ This shortage of affordable and available housing is not unique to New Haven. Nationally, there are only 24 available and affordable unassisted units for individuals earning less than 30% of the AMI.³⁷ Lack of affordable housing contributes to poor housing quality, cost burdens, eviction and instability, all of which have been linked to poor health. ^11–16

Given the shortage of private market rental housing, the unmet need for rental assistance may extend beyond existing waiting lists to virtually all low-income renters. The majority of participants in this study were neither on a waiting list nor receiving assistance, despite the fact that virtually all 400 participants were income eligible for rental assistance programs. While some individuals in this “neither” group may have acceptable and affordable unassisted housing, many may in fact stand to benefit from rental assistance yet face application or eligibility barriers. Comparisons between this neither group and those who are receiving assistance with respect to health are difficult to interpret due to potential unobserved heterogeneity. Factors that create barriers to applying for rental assistance, such as depression, anxiety, or other physical and mental health limitations, may also be associated with poor self-rated health. However, some barriers to assistance, such as closed waiting lists, may be exogenous to the individual. More research is needed to understand the extent to which unmet need for assistance may adversely affect the health of those who aren’t on housing waiting lists. An important starting point for this research is to more fully examine barriers to rental assistance applications.

Criminal justice involvement may be a particular barrier to applying for or receiving rental assistance. Qualitative research finds that individuals with recent incarceration histories often perceive themselves to be ineligible for this assistance.²⁵ Qualitative work also describes an arduous process that formerly incarcerated individuals can face when trying to prove themselves deserving of and eligible for rental assistance in a system that often defines eligibility on a case-by-case basis. ^24,25,38 We did not find that criminal justice variables (felony convictions or recent incarcerations) significantly modified the observed association between rental assistance and health, suggesting that this assistance is beneficial regardless of criminal justice history and may improve health even for those who face numerous barriers to well-being following a prison stay.²⁸ It is possible, however, that our study was not adequately powered to detect these interactions.

There are a few possible limitations to consider when interpreting our findings. First, though waiting lists provide a useful control group of individuals similar to those receiving assistance, there is a possibility of unobserved differences between these groups due to the possible prioritization of some households or due to potential eligibility barriers that could arise between the time of rental assistance application and receipt of housing. However, despite potential unobserved differences, our findings remain large and statistically significant even after controlling for several variables that could affect access to rental assistance. Furthermore, our findings are robust to multiple sensitivity analyses, including analyses that exclude homeless individuals from the unassisted groups.

Additionally, the cross-sectional nature of our data precludes our ability to infer causal effects of rental assistance on health. We cannot rule out the possibility of reverse causality: that poor health is itself a barrier to rental assistance. In particular, it is possible that mental health challenges create barriers to obtaining rental assistance. Though qualitative data collected as part of JustHouHS also suggest that mental health challenges could improve access to rental assistance when individuals with mental health needs are prioritized for housing or obtain rental assistance through supportive housing programs. Our disability measure captures some of the diagnosed mental illness in our sample. However, it does not capture undiagnosed or less severe mental illness. In order to consider the role of mental health, including undiagnosed illness, we conducted supplemental analyses (results not shown) that include a measure of psychological distress (Kessler’s K6 Scale)³⁹ as a control variable. In these models, group differences in self-rated health by rental assistance status remain large and statistically significant, suggesting that mental health symptoms do not explain our findings. Nonetheless, future longitudinal research that can examine transitions from waiting lists to rental assistance is needed to fully address the possibility of reverse causality.

An additional limitation is that our sample is not representative of rental assistance applicants and recipients in New Haven or beyond New Haven. Indeed, when we compare our sample with HUD data on recipients of HUD rental assistance in New Haven, we see that our sample is more male (40% vs approximately 20%) and less likely to have children in the household (17.8% vs 37%).³⁰ Furthermore, due to JustHouHS’s deliberate oversampling of those with recent incarcerations, our sample likely contains a larger portion of individuals with felony convictions or recent prison stays than the overall population of rent-assisted and waitlisted households in New Haven. In this sense, our sample may be particularly disadvantaged. Though our findings may not be generalizable, they do suggest that despite the numerous challenges that JustHouHS participants contended with, rental assistance mattered for their health and well-being.

Additionally, our sample size did not allow us to examine differences between types of rental assistance. Some previous research finds health benefits for project-based housing, but not voucher-based rental assistance.³ Understanding potential differences across types of rental assistance is important to inform future housing policy, particularly in light of policy shifts away from project-based housing toward tenant-based assistance in the form of vouchers.⁴⁰

Finally, as with any self-reported variable, there is possibility of misclassification of our rental assistance measure, though we attempt to minimize this misclassification by asking detailed and locally relevant questions about each form of rental assistance. Relatedly, it is possible that not all participants who receive voucher-based assistance are living in rent-assisted housing. The challenges that voucher holders can face in finding an eligible unit and a landlord who will rent to them are well documented, and it is possible that some vouchers are unused.⁴¹ However, this type of misclassification would attenuate any observed effects, making our findings conservative. Finally, we are unable to determine whether waitlisted individuals have access to other forms of affordable housing (for example, through family members, through tax credits that create affordable units, or through rent-restrictions). However, again, this possibility would attenuate group differences. Our findings suggest that even with the potential availability of other forms of affordable housing, rental assistance was still associated with improved health and as prior research suggests, likely contributes to improved housing access.¹⁰

Despite these limitations our study contributes to an emerging body of evidence on the health benefits of rental assistance and affordable housing more broadly. We extend this literature to consider the intersection of two national crises: a shortage of affordable housing and mass incarceration that creates additional barriers to housing and financial well-being for those released from prison. Though more research is needed, we find evidence suggesting that investments in rental assistance programs may improve the health of low-income Americans, perhaps reducing existing socioeconomic health disparities as well as health-care costs.