Housing Cost Burden and Health Decline Among Low- and Moderate-Income Older Renters

Abstract

Objective

To examine the association between housing cost burden (HCB) and health decline among low- and moderate-income older renters in the United States.

Method

Baseline data include low- and moderate-income community-dwelling older renters (N = 1,064) from the nationally representative 2015 National Health and Aging Trends Study. HCB was defined as the percentage of monthly income spent on rent, categorized as “no HCB” (<30%), “moderate HCB” (30%–49%), and “severe HCB” (≥50%). We used weighted logistic regression models to estimate whether HCB status in 2015 and change in HCB between 2015 and 2017 were associated with self-rated health decline and developing a new limitation related to activities of daily living (ADL) or instrumental activities of daily living (IADL) between 2015 and 2017.

Results

Older renters with severe HCB in 2015 were the most likely to develop a new ADL/IADL limitation (63.4%) over time (p < .05). The association between HCB status in 2015 and self-rated health decline was not statistically significant, but older renters with persistent HCB had 1.64 times greater odds of self-rated health decline (p < .05) and 2.01 times greater odds of developing a new ADL/IADL limitation (p < .01), compared to older renters with no HCB at baseline and follow-up.

Discussion

Even in the short term, HCB contributes to health decline in later life. Efforts to promote equity and healthy aging in the community must consider how to best address housing affordability among the growing population of older renters.

Activities of daily living, Affordable housing, Health-related social needs, Healthy aging

Coinciding with population aging and the growing affordable housing crisis, older Americans are increasingly spending more of their income on housing (Joint Center for Housing Studies, 2019) and often face trade-offs between covering housing costs or paying for food, medications, and other health-related expenses (Alley et al., 2011; Pollack et al., 2010). The number of older adults experiencing housing cost burden (HCB), commonly defined as spending more than 30% of household income on housing costs, has reached an all-time high with roughly 10 million older households considered burdened by high housing expenses (Joint Center for Housing Studies, 2019). In the United States, the availability of publicly funded housing assistance falls substantially short of demand with only 33% of eligible low-income older households receiving federal assistance (U.S. Department of Housing and Urban Development [HUD], 2017). Older adults in need of housing assistance remain on waitlists for months or even years (HUD, 2019) and often experience food insecurity, housing instability, and poor health outcomes while waiting for assistance (Carder et al., 2018).

Roughly a third of older households experience HCB, with older renters more likely to face unaffordable housing costs compared to older homeowners (54% vs 26%; Joint Center for Housing Studies, 2019). Although the homeownership rate among older adults (79%) is generally high compared to younger populations (U.S. Census Bureau, 2018), evidence suggests that homeownership rates are declining as rents are rising for older Americans (Joint Center for Housing Studies, 2019). The growing proportion of older renters contributes to financial insecurity in older adulthood because home equity is the largest source of retirement savings for most older Americans (Sass, 2017). Older renters also experience less stable housing costs and substantially lower levels of household wealth compared to older homeowners (Joint Center for Housing Studies, 2019).

Background

Given the current affordable housing crisis and declining homeownership rates among older adults (Joint Center for Housing Studies, 2019), it is important to better understand the potential health consequences associated with living in unaffordable housing among the growing population of older renters. Research suggests that living in unaffordable housing is associated with poor health outcomes (Meltzer & Schwartz, 2015; Pollack et al., 2010) and that receiving subsidized housing can improve self-rated health (Fenelon et al., 2017). However, there is limited research examining the association between housing affordability and health among older renters, with some studies specifically excluding older adults due to their unique needs (Bentley et al., 2011; Simon et al., 2017).

The research designs of studies that do focus on the older population often make it challenging to understand the independent contribution of housing affordability on health. For instance, various studies have examined the impact of living in affordable senior housing on healthy aging outcomes, but these studies either test the impact of a specific supportive service intervention (Castle, 2008; Freeman et al., 2018; Kandilov et al., 2018) or it is not possible to disentangle the effects of the supportive services intervention from the effects of living in affordable housing (Fonda et al., 2002; Gusmano et al., 2018). Recent studies using nationally representative data from the Health and Retirement Study (HRS) also provide evidence that use of senior housing for low- and moderate-income older adults promotes a variety of healthy aging outcomes (Kim et al., 2018; Park et al., 2017, 2018), but less is known about the health benefits of living in affordable housing outside of congregate senior housing arrangements.

An earlier study using a sample of community-dwelling older adults from the HRS examined the effect of housing disadvantage (renter, low-quality housing, unaffordable housing, or low neighborhood safety) on health decline in later life (Alley et al., 2009). The results suggest that housing disadvantage independently contributes to self-rated health decline among community-dwelling older adults. Similarly, a recent study using the HRS found that older adults experiencing financial strain and living in poor-quality housing are more likely to have higher levels of cardiometabolic risk, even after accounting for socioeconomic disadvantage (Mawhorter et al., 2021). However, because housing affordability was not the construct of interest in these studies, the results do not adequately inform our understanding of the association between HCB and health decline among older renters.

The Current Study and Conceptual Framework

Although previous studies collectively suggest that access to affordable housing promotes healthy aging in the community, the independent contribution of housing affordability on health among older renters is not well understood. The current prospective study addresses gaps in the literature by examining if experiencing moderate or severe HCB (compared to no HCB) in 2015 contributes to health decline by 2017 among low- and moderate-income older renters. We also examine whether change in HCB status between baseline and follow-up is associated with health decline during the same 2-year period. The results of this study can inform programs and policies aimed at addressing the social and economic determinants of older adults’ health by improving our understanding of the housing and health connection in later life.

Our conceptual framework is informed by the work of Taylor (2018), who describes housing as a key social determinant of health that operates through four primary pathways: the stability pathway (being homeless or experiencing housing instability), quality/safety pathway (poor conditions of the home), affordability pathway (financial burden from high housing costs), and neighborhood pathway (environmental/social characteristics of where people live) (Taylor, 2018). Taken together, these four pathways (stability, quality, affordability, and neighborhood) represent the larger concept of housing security. Of the four housing pathways, the affordability pathway has the smallest evidence base linking affordable housing to health (Maqbool et al., 2015). The current study focuses on identifying this potential pathway but does not test the specific mechanisms that link affordability to health decline. Yet previous work suggests that increased stress and/or financial trade-offs help explain this association. Evidence suggests that housing insecurity contributes to elevated stress levels (Singh et al., 2019) and creates trade-offs between covering housing expenses and paying for health-related goods and services (e.g., nutritious food, medications, preventive health services; Baker et al., 2014; Meltzer & Schwartz, 2015).

Our conceptual framework also acknowledges that although the housing pathways are distinct, they are also interrelated (Mason et al., 2013; Pollack et al., 2010), so when examining the affordability pathway, it is essential to consider the quality, neighborhood, and stability pathways simultaneously. For example, living in a home in need of significant repairs or in a neighborhood with few health-promoting features is likely to be more affordable, but the health benefits associated with having affordable housing costs may be offset by poor home quality and/or poor neighborhood conditions. Indeed, previous research on young families suggests that low housing costs are not always better and can negatively affect children’s cognitive achievement because it is likely associated with poor-quality housing or neighborhood (Newman & Holupka, 2016). Therefore, to assess the independent contribution of the affordability pathway on health, we control for measures of home and neighborhood quality in the current study. Living in unaffordable housing can also lead to housing instability (Desmond, 2015), which can make the affordability and stability pathways hard to disentangle (Taylor, 2018). In this way, the association between unaffordable housing and health decline may be partially mediated by experiencing housing instability. Unfortunately, this mediation effect is challenging to assess in longitudinal studies because older adults experiencing housing instability are less likely to remain in the study over time. For this reason, in the current study, we explicitly examine the association between HCB and attrition to better understand our final estimates. We expect that older renters experiencing HCB are more likely to be lost to attrition, compared to renters with affordable housing costs.

This study uses longitudinal data from the National Health and Aging Trends Study (NHATS) and is the first nationally representative study we know of to examine if experiencing HCB contributes to health decline among older renters. We focus on low- and moderate-income older renters to test the following hypotheses:

• Hypothesis 1: Older renters with severe HCB at baseline (≥50% of monthly income spent on rent) will be most likely to experience health decline over time and older renters with moderate HCB (30%–49% of monthly income spent on rent) at baseline will be more likely to experience health decline 2 years later compared to older renters with no HCB (<30% of income spent on rent).

• Hypothesis 2: Older renters with persistent HCB (moderate or severe HCB at baseline and follow-up) will be most likely to experience health decline compared to older renters who continue to have no HCB at baseline and follow-up, those with resolved HCB (moderate or severe HCB at baseline, but no HCB at follow-up), and those with developed HCB (no HCB at baseline, but moderate or severe HCB at follow-up).

Method

Data

We use baseline data from the 2015 wave of the NHATS and follow-up data from 2017 to examine both self-rated health decline and developing a new limitation related to activities of daily living (ADL) or instrumental activities of daily living (IADL) over 2 years. Conducted annually since 2011 and replenished in 2015 to adjust for attrition, the NHATS is a nationally representative sample of Medicare beneficiaries aged 65 and older. More information on the NHATS sampling design and methods can be found in detail elsewhere (DeMatteis et al., 2016b).

We first limit our analytic sample to renters living in the community in 2015 (n = 1,244), excluding residents of Continuing Care Retirement Communities, assisted living, nursing homes, and other arrangements (e.g., living with family). Consistent with prior work (Jenkins Morales & Robert, 2020), we exclude older renters in the top 25% of the income distribution with incomes more than $65,000 per year to limit our sample to low- and moderate-income older renters (n = 1,161). Because the association between HCB and health decline is likely contingent on limited disposable income, higher-income older adults are less likely to face the same financial trade-offs due to HCB and are less likely to have HCB, compared to low- and moderate-income older adults. As confirmation of this, among older renters in the 2015 NHATS sample, less than 10% of higher-income renters had moderate or severe HCB (n = 11), compared to 59% of low- and moderate-income older renters (n = 616).

If participants had missing data on covariates of interest, data from the subsequent round were used when available (n = 87). If Medicaid receipt was still unknown, we coded the participant as having Medicaid if their annual income was less than $15,000 (n = 9). Remaining cases with missing data on covariates of interest were deleted (n = 97). The final analytic sample consisted of 1,064 low- and moderate-income older renters in 2015.

We attempted to preserve the analytic sample as much as possible to test our first hypothesis that older renters with severe HCB at baseline will be most likely to experience health decline 2 years later. As such, if a participant was lost to attrition, died, or had missing data on the 2017 health measure, then data from the corresponding 2016 self-rated health (n = 135) or ADL/IADL limitation (n = 115) measure were used when possible. Participants who died (n = 45) or were lost to attrition by 2016 (n = 78) were excluded from the analysis. This process yielded complete follow-up data for 938 participants for self-rated health and 939 participants for the ADL/IADL limitation analyses.

To conduct the analysis for our second hypothesis, examining if older renters with persistent HCB were most likely to experience health decline over time, a subsample of low- and moderate-income older adults who were renters in 2015 and 2017 (n = 687) was used. Therefore, this analysis excluded any participants who died (n = 112), were lost to attrition (n = 141), were no longer a renter (n = 113), or had missing data on rental status in 2017 (n = 11). Table 2 and Supplementary Tables 1 and 2 examine differences between the analytic sample and the subsamples used to test our hypotheses. We discuss how the samples differ, specifically by HCB status, in the “Results” section.

Table 2.

Weighted Baseline Self-Rated Health, ADL/IADL Limitation, and Change at Follow-Up Among Low- and Moderate-Income Older Renters by HCB Status in 2015

	Totala sample	No HCB	Moderate HCB	Severe HCB	p
Baseline health in 2015
Self-rated health (1–5)	2.9	2.9	3.0	2.8	.189
Poor self-rated health (%)	10.0	9.0	9.4	11.8	.535
ADL/IADL limitation (0–12)	2.0	2.1	2.0	2.0	.889
Health change between 2015 and 2017 (%)
Self-rated health decline	43.9	43.6	45.7	42.6	.810
New ADL/IADL limitation	56.8	51.3	58.3	63.4	.015
Sample change between 2015 and 2016 (%)					.001
Alive and not lost to attrition by 2016	86.4	91.8	84.2	81.4
Deceased by 2016	4.0	3.6	6.2	2.4
Lost to attrition by 2016	9.6	4.5	9.6	16.2
Sample change between 2015 and 2017 (%)					.005
Alive and not lost to attrition by 2017	63.1	70.3	58.9	57.3
Deceased by 2017	9.7	8.7	12.8	8.3
Lost to attrition by 2017	17.7	11.3	19.6	24.5
No longer a renter by 2017	9.5	9.7	8.7	9.9

Notes: ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living. The mean is presented for continuous variables. The p value reflects the design-based F test that corrects for the National Health and Aging Trends Study complex survey design.

^aFor baseline health measures, sample change between 2015 and 2016, and sample change between 2015 and 2017 (N = 1,064). n = 861 for the self-rated health decline analysis, after excluding 77 participants who had poor health at baseline in 2015, and n = 939 for the new ADL/IADL limitation analysis.

Measures

Health outcomes

We include measures of both self-rated health decline and developing a new ADL/IADL limitation as the health outcomes of interest. We chose to focus on self-rated health (1 = poor; 5 = excellent), because it is often used in the literature as a general health measure to capture the participant’s view of their physical, social, emotional, psychological, and overall well-being and allows us to compare our results to prior work on housing affordability and health decline (Alley et al., 2009, 2011). We categorized a participant as experiencing self-rated health decline if their level of self-rated health decreased between baseline and follow-up (1= self-rated health decline).

To compliment the self-rated health measure, we chose ADL/IADL limitation as our other health outcome of interest, which is consistent with the World Health Organization’s (2020) focus on maintaining and improving functional ability as the primary target of their new worldwide initiative “Decade of Healthy Ageing 2021–2030.” This measure has not been used in prior work on housing affordability and health and will provide new insight on the potential connection between housing affordability and independent living capacity. To assess ADL/IADL limitation, NHATS participants were asked if they had difficulty completing 12 self-care, mobility, or household activities independently in the last month. ADL tasks included eating, toileting, bathing, dressing, getting around inside the home, and getting out of bed (0–12 scale). IADL tasks included doing laundry, shopping, preparing meals, handling bills/banking, getting outside the home, and managing medications. Participants who never completed a task independently, or received help for health or functioning reasons, were also coded as having a limitation. We categorized a participant as developing a new ADL/IADL limitation if the number of reported ADL/IADL limitations increased between baseline and follow-up (1 = new ADL/IADL limitation).

Housing cost burden

Two measures of HCB were used: HCB status in 2015 (hereafter referred to as HCB status) and change in HCB between 2015 and 2017 (hereafter referred to HCB change). To create the HCB measures, we first had to calculate the percentage of income spent on rent. To do so, we used self-reported monthly rent and the NHATS publicly available imputed income measure (based on a self-reported estimate of individual monthly income or income for coresiding couples [DeMatteis et al., 2016a]). Consistent with common thresholds of HCB, the 2015 HCB status measure consisted of three categories: no HCB (<30% of monthly income spent on rent), moderate HCB (30%–49% of monthly income spent on rent), and severe HCB (≥50% of monthly income spent on rent; Joint Center for Housing Studies, 2019). The HCB change measure consisted of four categories: consistent no HCB (those who continued to have no HCB in 2015 and 2017), resolved HCB (moderate or severe HCB in 2015, but no HCB in 2017), developed HCB (no HCB in 2015, but moderate or severe HCB in 2017), and persistent HCB (moderate or severe HCB in 2015 and 2017).

Sociodemographic characteristics

We include self-reported gender (female = 1; male = 0), age categorized ordinally from 1 to 6 (1 = 65–69 years old; 6 = 90+ years old), race/ethnicity (non-Hispanic White, non-Hispanic Black, or Hispanic; another race/ethnicity), living arrangement (lives alone = 1; lives with others = 0), education (below high school, high school, beyond high school), Medicaid receipt (Medicaid = 1), and imputed total annual income (log-transformed to better approximate the nonlinear relationship between income and health). The NHATS publicly available imputed income measure has been described in detail elsewhere (DeMatteis et al., 2016a) and is based on a self-reported estimate of individual monthly income or income for coresiding couples from various sources, including earned income, social security payments, pensions, supplemental security income, retirement account withdrawals, and income from interest/dividends (Kasper & Freedman, 2018).

Housing context

When trying to isolate the potential impact of housing affordability on health decline, we control for other housing measures that are interrelated and might contribute to health in later life. We control for housing quality because affordability and quality are often negatively correlated (Spillman et al., 2012). Any interviewer-observed housing quality concern (e.g., paint peeling inside the home, pests, broken furniture, flooring problems, broken windows, crumbling foundation, siding issues, roof problems, or broken stairs) was used to create a dichotomous measure of housing quality (1 = housing quality concern).

The shortage of affordable housing stock often relegates low-income households to neighborhoods with higher rates of poverty and fewer resources for health promotion (outdoor recreation, access to quality health services, etc.), so we also control for neighborhood factors when assessing the association between affordable housing and health decline (Braveman et al., 2011). Neighborhood social cohesion was measured based on participants’ responses to three questions: (1) people in my community know each other well, (2) people are willing to help each other, and (3) people can be trusted (α = .74). Community was self-defined by the participant and each question was answered on a 3-point scale (1 = do not agree; 2 = agree a little; 3 = agree a lot). Similar to previous studies (Latham & Clarke, 2018; Millar, 2019), we created an average composite score for each participant and identified scores in the lowest 15th percentile of the distribution (scores ≤ 1.67) to generate a dichotomous indicator of “low neighborhood cohesion.” Neighborhood physical disorder, based on interviewer observation, was also included as a control to capture the housing context of participants. Interviewers completed an environmental checklist and rated the level (1 = none, 2 = a little, 3 = some, 4 = a lot) of (1) litter/trash on sidewalks, (2) graffiti/vandalism, and (3) vacant homes/stores in the area, to determine neighborhood physical disorder. Because interviewer observation of neighborhood physical disorder was rare (<11% of participants), we created a dichotomous measure (1 = neighborhood physical disorder) to capture any sign of disorder (Latham & Clarke, 2018).

Analysis

First, bivariate analyses were conducted to examine differences in baseline sociodemographic characteristics, housing context, and health by HCB status in 2015. To better understand our final estimates, we used the chi-squared test and multinomial logistic regression models (Supplementary Tables 1 and 2) to examine if participants experiencing HCB at baseline were more likely to die, be lost to attrition, or no longer be a renter (e.g., moved in with family) over time, compared to those without HCB.

To address our first hypothesis, we used the chi-squared test to examine if experiencing self-rated health decline or developing a new ADL/IADL limitation between 2015 and 2017 varied by HCB status in 2015 (n = 939). We then used logistic regression models to examine the odds of experiencing self-rated health decline or developing a new ADL/IADL limitation by 2015 HCB status, before and after adjusting for covariates. By definition, a decline in self-rated health or developing a new ADL/IADL limitation cannot be observed for participants who reported poor self-rated health (n = 77) or a limitation with all 12 ADL/IADL tasks (n = 0) at baseline. For this reason, 77 participants were excluded from the self-rated health decline analysis (n = 861).

To address our second hypothesis, we used a subsample of low- and moderate-income older adults who were renters in 2015 and 2017 (n = 687) to examine if older renters with persistent HCB were more likely to experience health decline compared to those with consistent no HCB, resolved HCB, and developed HCB between 2015 and 2017. Consistent with our analytic strategy for our first hypothesis, participants with poor self-rated health at baseline (n = 60) were also excluded from the HCB change analysis (n = 627). Bivariate analyses were used to describe the baseline sociodemographic characteristics and housing context of the subsample and to examine if experiencing self-rated health decline or developing a new ADL/IADL limitation varied by HCB change between 2015 and 2017. We then used logistic regression models to examine the odds of experiencing self-rated health decline or developing a new ADL/IADL limitation by HCB change, before and after adjusting for covariates. All analyses were conducted in Stata version 16.0 (Stata Corp, College Station, TX). The NHATS 2015 analytic weights were used in all analyses and standard errors were corrected based on the NHATS’ complex sample design (Kasper & Freedman, 2018).

Results

Table 1 presents weighted characteristics of the sample by HCB status in 2015. The majority (59.1%) of low- and moderate-income older renters in our analytic sample had moderate (27.6%) or severe (31.5%) HCB in 2015, which is comparable to other national estimates (Joint Center for Housing Studies, 2019). The average age of the analytic sample was roughly 75 years old. Older renters with severe HCB were more likely to be women (67.9%) compared to those with no (56.8%) or moderate (61.3%) HCB (p = .055). Non-Hispanic White older renters were less likely to have moderate or severe HCB, comprising 60.7% of the analytic sample and 56.6% of those with moderate or severe HCB (p = .085). Older renters with severe HCB were most likely to live with others (55.9%; p < .001). Although educational attainment and Medicaid receipt were not significantly different by HCB status, older renters with severe HCB had significantly lower average annual income ($14,170) compared to those with moderate ($21,571) or no ($26,069) HCB (p < .001). There were no statistically significant differences in observed housing quality and neighborhood physical disorder across HCB status levels. However, older renters with moderate (28.1%) or severe (31.7%) HCB were more likely to self-report living in a neighborhood characterized by low social cohesion compared to older renters with no HCB (19.6%; p = .021). As seen in Table 2, weighted baseline self-rated health and the average number of ADL/IADL limitations were similar for older renters with no, moderate, and severe HCB in 2015. On average, participants rated their health as “good” and reported two ADL or IADL limitations. Roughly 10% of the analytic sample rated their health as “poor” at baseline.

Table 1.

Weighted Baseline Characteristics of Low- and Moderate-Income Older Renters by HCB Status in 2015

	Total sample	No HCB	Moderate HCB	Severe HCB	p
2015 Sociodemographic characteristics
Female (%)	61.5	56.8	61.3	67.9	.055
Age (1–6 scale)	2.5	2.5	2.5	2.6	.923
Race/ethnicity (%)					.085
Non-Hispanic White	60.7	65.3	58.2	56.9
Non-Hispanic Black	16.0	16.3	15.1	16.4
Hispanic	16.6	15.0	17.0	18.4
Another race/ethnicity	6.7	3.4	10.0	8.3
Lives alone (%)	53.1	61.5	50.7	44.1	<.001
Education (%)					.580
Below high school	34.0	32.8	32.4	37.0
High school	27.3	27.8	25.4	28.2
Beyond high school	38.7	39.4	42.2	34.8
Average annual income (dollars)	21,083	26,069	21,571	14,170	<.001
Medicaid	39.4	39.4	33.6	44.5	.129
Housing quality concern (%)	21.9	24.2	20.1	20.5	.597
Neighborhood physical disorder (%)	17.3	15.7	14.8	21.6	.121
Low neighborhood social cohesion (%)	25.7	19.6	28.1	31.7	.021
Total sample (%)		41.0	27.6	31.5

Notes: HCB = housing cost burden. The mean is presented for continuous variables. No HCB (<30% of income spent on rent); moderate HCB (30%–49% of income spent on rent); severe HCB (≥50% of income spent on rent). N = 1,064. The p value reflects the design-based F test that corrects for the National Health and Aging Trends Study complex survey design.

Before focusing on our main results, we report results of chi-squared tests to examine if older renters with moderate or severe HCB were more likely to die, be lost to attrition, or no longer be a renter (e.g., move in with family or to a residential facility) between baseline and follow-up. As seen in Table 2, there was no statistically significant difference in remaining a renter in the community between 2015 and 2017 by HCB status. Because older renters who died or were lost to attrition between 2015 and 2016 were excluded from all longitudinal analyses, we report those results here; however, the same patterns emerged when examining death and attrition between 2015 and 2017. As seen in Table 2, older renters with severe HCB (16.2%) and moderate HCB (9.6%) were more likely to be lost to attrition between 2015 and 2016 than older renters with no HCB (4.5%) in 2015. Compared to older renters with no HCB (3.6%) and severe HCB (2.4%), older renters with moderate HCB (6.2%) were most likely to die by 2016. However, the number of participants with severe HCB who died by 2016 is likely underestimated due to higher attrition rates. Supplementary Tables 1 and 2 also suggest that even after controlling for baseline sociodemographic characteristics, older renters with HCB were more likely to die or be lost to attrition over time. Collectively, these results show that, as we would expect, older renters with HCB are underrepresented in our longitudinal analyses due to death and attrition, suggesting that our final estimates of the impact of HCB on health decline will be underestimated.

Health Decline and HCB Status in 2015

Our first hypothesis is that older renters with severe HCB at baseline in 2015 will be most likely to experience health decline over time, and that older renters with moderate HCB will be more likely to experience health decline compared to older renters with no HCB. As seen in Table 2, contrary to our first hypothesis, older renters with no (43.6%), moderate (45.7%), and severe (42.6%) HCB in 2015 experienced similar rates of self-rated health decline over two years (p = .810). In the multivariate analysis (Table 3), HCB status in 2015 was still not associated with self-rated health decline even after adjusting for differences in baseline sociodemographic characteristics and housing context.

Table 3.

Relative Odds of Self-Rated Health Decline and New ADL/IADL Limitation Between 2015 and 2017 by HCB Status in 2015 Among Low- and Moderate-Income Older Renters

	Self-rated health decline		New ADL/IADL limitation
	Unadjusted	Adjusted	Unadjusted	Adjusted
HCB (ref. = no HCB)
Moderate HCB	1.09	1.06	1.32	1.46*
Severe HCB	0.96	1.01	1.64**	1.94**
Female		0.83		1.00
Age (1–6 scale)		1.00		1.09
Race/ethnicity (ref. = non-Hispanic White)
Non-Hispanic Black		0.81		1.03
Hispanic		1.55		1.31
Another race/ethnicity		1.98		0.78
Lives alone		0.88		1.51*
Education (ref. = below high school)
High school		1.49†		1.21
Beyond high school		1.09		0.88
Annual income (logged)		1.12		1.14†
Medicaid		1.34		1.35†
Housing quality issue		0.82		1.18
Neighborhood physical disorder		0.81		0.83
Low neighborhood social cohesion		0.99		1.09
Constant	0.77*	0.24	1.05	0.15*

Notes: ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living. Results are presented as odds ratios. Regressions employ weights and adjust for standard errors to account for complex survey design. Participants who had poor self-rated health (n = 77) at baseline in 2015 were excluded. n = 861 for the self-rated health decline analysis and n = 939 for the new ADL/IADL limitation analysis.

^†p < .10. *p < .05. **p < .01. ***p < .001.

However, in support of our first hypothesis, older renters with severe HCB were the most likely to develop a new ADL/IADL limitation (63.4%) between baseline and follow-up, compared to those with moderate (58.3%) or no (51.3%) HCB in 2015 (p = .015). In the adjusted model (Table 3), older renters with severe HCB in 2015 had 1.94 times greater odds (odds ratio [OR] = 1.94, SE = 0.38) of developing a new ADL/IADL limitation over 2 years compared to older renters with no HCB (p < .01). Older renters with moderate HCB were also more likely (OR = 1.46, SE = 0.27) to develop a new ADL/IADL limitation compared to those with no HCB in 2015 (p < .05).

Health Decline and HCB Change Between 2015 and 2017

To test our second hypothesis that older renters with persistent HCB will be most likely to experience health decline compared to older renters with consistent no HCB, resolved HCB, and developed HCB between baseline and follow-up, we used a subsample of low- and moderate-income older adults who were renters in both 2015 and 2017 (n = 687). Table 4 presents weighted characteristics of the subsample by HCB change. The majority (75.3%) of the subsample had consistent no HCB (32.7%) or persistent HCB (42.6%) with roughly a quarter of the subsample experiencing a change in HCB status between 2015 and 2017. Baseline self-rated health and the average number of ADL/IADL limitations were similar for older renters by HCB change. As with the larger sample in the previous analysis, on average, participants in this subsample rated their health as “good” and reported roughly two ADL or IADL limitations.

Table 4.

Weighted Baseline Characteristics and Health Change at Follow-Up by Change in HCB Status Between 2015 and 2017 Among Low- and Moderate-Income Older Renters

	Total sample	Consistent No HCB	Resolved HCB	Developed HCB	Persistent HCB	p
2015 Sociodemographic characteristics
Female (%)	59.2	54.6	60.6	51.0	64.8	.108
Age (1–6 scale)	2.5	2.3	2.5	2.6	2.6	.289
Race/ethnicity (%)						.078
Non-Hispanic White	60.0	67.6	59.6	56.5	55.2
Non-Hispanic Black	18.1	17.3	24.0	19.8	16.7
Hispanic	16.7	11.6	12.7	22.7	19.8
Another race/ethnicity	5.3	3.5	3.7	1.0	8.3
Lives alone (%)	56.1	58.3	58.1	68.5	50.1	.124
Education (%)						.133
Below high school	34.1	27.5	28.6	43.4	37.9
High school	25.9	28.5	19.6	26.8	25.4
Beyond high school	40.0	44.0	51.9	29.8	36.7
Average annual income (dollars)	21,161	26,840	15,805	24,667	17,180	<.001
Medicaid	40.8	38.7	32.1	49.0	42.2	.305
Housing quality concern (%)	22.1	24.7	16.8	21.3	21.7	.714
Neighborhood physical disorder (%)	19.5	15.5	13.1	19.8	24.3	.102
Low neighborhood social cohesion (%)	23.9	18.7	24.5	27.3	26.6	.471
2015 Health measures
Self-rated health (1–5)	2.9	3.0	2.9	2.8	2.9	.539
Poor self-rated health (%)	8.6	5.4	11.8	12.5	8.9	.291
ADL/IADL limitation (0–12)	1.8	1.6	1.8	2.5	1.9	.162
Change between 2015 and 2017 (%)
Self-rated health declinea	47.3	44.4	29.4	49.8	53.6	.018
New ADL/IADL limitation	57.7	51.6	51.6	51.2	66.0	.012
Total sample (%)		32.7	11.6	13.2	42.6

Notes: ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living. The mean is presented for continuous variables. n = 687. The p value reflects the design-based F test that corrects for the National Health and Aging Trends complex survey design.

^aParticipants who had poor self-rated health (n = 60) at baseline in 2015 were excluded.

In support of our second hypothesis, as seen in Table 4, older renters with persistent HCB were most likely to experience self-rated health decline (53.6%) compared to those with consistent no HCB (44.4%), resolved HCB (29.4%), and developed HCB (49.8%) (p = .018). In the adjusted model (Table 5), older renters with persistent HCB had 1.64 times greater odds (OR = 1.64, SE = 0.34) of experiencing self-rated health decline over 2 years compared to older renters with consistent no HCB (p < .05). There was no statistically significant difference between older renters with persistent HCB and those who developed HCB between baseline and follow-up. However, those with persistent HCB were much more likely (OR = 2.58, SE = 0.78) to experience self-rated health decline compared to those with resolved HCB (p < .01; results not shown).

Table 5.

Relative Odds of Self-Rated Health Decline and New ADL/IADL Limitation Between 2015 and 2017 by HCB Change Between 2015 and 2017 Among Low- and Moderate-Income Older Renters

	Self-rated health decline		New ADL/IADL limitation
	Unadjusted	Adjusted	Unadjusted	Adjusted
HCB (ref. = consistent no HCB)
Resolved HCB	0.52*	0.63	1.00	1.16
Developed HCB	1.24	1.35	0.98	0.87
Persistent HCB	1.45†	1.64*	1.82**	2.01**
Female		0.75		0.93
Age (1–6 scale)		0.96		0.99
Race/ethnicity (ref. = non-Hispanic White)
Non-Hispanic Black		0.64†		0.89
Hispanic		1.52		1.47
Another race/ethnicity		1.12		0.68
Lives alone		0.73		1.62†
Education (ref. = below high school)
High school		1.71*		1.40
Beyond high school		1.11		0.93
Annual income (logged)		1.14		1.10
Medicaid		1.54†		1.28
Housing quality issue		0.87		1.40
Neighborhood physical disorder		0.62		0.82
Low neighborhood social cohesion		0.92		1.40
Constant	0.80	0.26	1.07	0.24

Notes: ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living. Results are presented as odds ratios. Regressions employ weights and adjust for standard errors to account for complex survey design. Participants who had poor self-rated health (n = 60) at baseline in 2015 were excluded. n = 627 for self-rated health decline analysis and n = 687 for new ADL/IADL limitation analysis.

^†p < .10. *p < .05. **p < .01. ***p < .001.

The ADL/IADL results also suggest support for our second hypothesis. As seen in Table 4, older renters with persistent HCB (66.0%) were most likely to develop a new ADL/IADL limitation between 2015 and 2017 compared to those with consistent no HCB (51.6%), resolved HCB (51.6%), and developed HCB (51.2%) (p = .012). In the adjusted model (Table 5), older renters with persistent HCB had 2.01 times greater odds (OR = 2.01, SE = 0.42) of developing a new ADL/IADL limitation over 2 years compared to older renters with consistent no HCB (p < .01). Those with persistent HCB were also significantly more likely to develop an ADL/IADL limitation compared to older renters with developed (OR = 2.32, SE = 0.69; p < .01) or resolved (OR = 1.74, SE = 0.51; p < .10) HCB between 2015 and 2017 (results not shown).

Discussion

Overall, the results support our hypotheses and suggest that experiencing HCB contributes to health decline among low- and moderate-income older renters and that experiencing persistent HCB over time is associated with increased risk of health consequences associated with unaffordable housing costs. The results of this study build on prior work in several ways. This is the first study we know of to examine the association between HCB and the development of new ADL/IADL limitations and the results suggest that experiencing both moderate and severe HCB (compared to no HCB) may contribute to increasing functional limitations, thus affecting independent living capacity among older adults. This finding underscores the importance of the connection between affordable housing and long-term care policy. Research on the association between housing affordability and self-rated health decline among older adults is sparse and the findings from prior studies are mixed (Alley et al., 2009, 2011). Although we do not find a statistically significant association between HCB in 2015 and self-rated health decline 2 years later, a novel finding from this study is that persistent HCB may contribute to self-rated health decline among older adults.

A strength of our study is that we attempted to isolate the effects of HCB on health by controlling for other aspects of housing that might impact health—self-reported and interviewer-observed neighborhood characteristics and housing quality measures. Consistent with prior work (Newman & Holupka, 2016), the results also suggest that experiencing HCB independently contributes to health decline and is not simply explained by lower incomes among older renters with moderate or severe HCB. Although interventions designed to increase income among older adults can in turn reduce HCB, targeted programs that specifically reduce HCB, such as HUD rental assistance, should also be considered health-related social interventions that have the potential to promote healthy aging in the community. The results of this study can help housing advocates gain support from policy makers to invest in funding for affordable housing because evidence suggests that understanding the connection between housing and health can build political support for affordable housing interventions (Ortiz et al., 2020). In addition to improved access to HUD rental assistance, developing and expanding creative solutions to reduce housing costs, such as elder cohousing models, are also needed. Given the variety of needs and preferences among older adults, having access to a variety of affordable housing options, regardless of the health benefits, should also motivate future research and policy.

Considering that renters with no HCB were disproportionately non-Hispanic White, the results also suggest that structurally rooted housing disparities might contribute to known racial disparities in healthy aging (Ferraro et al., 2017). The prevalence of multigenerational households also varies by race and ethnicity (Joint Center for Housing Studies, 2019), which could influence the association between housing affordability and health among older adults. Due to sample size constraints, we were unable to examine within and between racial/ethnic differences in the current study. Future research should examine how unequal access to housing security contributes to structurally rooted health inequities among older adults (Swope & Hernández, 2019).

More research is also needed to understand the pathways connecting HCB to health decline to effectively target potential interventions. Based on prior research (Baker et al., 2014; Singh et al., 2019), we expect that the association between HCB and health decline is explained by a material (e.g., financial trade-offs) and/or stress pathway, but we did not test these specific mechanisms. Prior research also affirms the interconnectedness of material resources on health among older adults (Alley et al., 2009), but we do not know how these connections might impact our results. For example, experiencing HCB could contribute to food insecurity, which in turn could lead to health decline. It is also plausible that financial trade-offs between covering housing or medical expenses helps explain the association between HCB and health decline among older renters over the short follow-up period. Older adults with HCB were also more likely to drop from the study between baseline and follow-up, suggesting that housing instability might mediate the association between HCB and health decline.

Even though we use longitudinal data and found significant associations between HCB and health decline, we cannot conclude with certainty that HCB causes health decline among older adults. Evidence suggests that the association between affordable housing and health is bidirectional—living in unaffordable housing contributes to poor health and people experiencing health problems are more likely to have difficulty paying their rent or mortgage (Baker et al., 2014). In addition, selection bias might occur if older adults chose housing arrangements based on unmeasured information that affects their self-assessments of future health. While considering these limitations, our findings suggest that, even in the short term, living in unaffordable housing may lead to health consequences among older renters. Our final estimates are also likely downwardly biased because older renters with HCB are underrepresented in our longitudinal analyses due to death and attrition. It is also possible that our results are downwardly biased because our measure of HCB does not include the cost of utilities (in addition to rent) as part of housing costs, and therefore likely underestimates the number of older adults who might be considered cost burdened (Airgood-Obrycki et al., 2021).

The results of this study provide additional evidence that aging in the community is a greater challenge for renters with unaffordable housing costs (Jenkins Morales & Robert, 2020) with tangible effects on their health. Given the growing population of older renters with HCB in the United States (Joint Center for Housing Studies, 2019), policy makers and program administrators should consider how the affordable housing crisis might undermine national efforts to promote healthy aging in the community.

Funding

This work was supported by the National Institutes of Health funding to the University of Wisconsin–Madison Center for Demography of Health and Aging (P30 AG017266), and the University of Wisconsin–Madison Office of the Vice Chancellor for Research and Graduate Education with funding from the Wisconsin Alumni Research Foundation. The National Health and Aging Trends Study (NHATS) is sponsored by the National Institute on Aging (NIA U01AG032947) through a cooperative agreement with the Johns Hopkins Bloomberg School of Public Health.

Conflict of Interest

None declared.

Author Contributions

M. Jenkins Morales and S. A. Robert planned the study and contributed to writing the article. M. Jenkins Morales performed the statistical analysis in consultation with S. A. Robert.