Week 4 – Assignment 1: Examine Executive Visioning and Week 5 - Assignment: Assess Leadership Challenges

Housing Assistance Programs and Adult Health in the United States

Andrew Fenelon, PhD, Patrick Mayne, MPH, Alan E. Simon, MD, Lauren M. Rossen, PhD, Veronica Helms, MPH, Patricia Lloyd, PhD, Jon Sperling, PhD, and Barry L. Steffen, MS

Objectives. To examine whether access to housing assistance is associated with better

health among low-income adults.

Methods. We used National Health Interview Survey data (1999–2012) linked to US

Department of Housing and Urban Development (HUD) administrative records (1999–

2014) to examine differences in reported fair or poor health and psychological distress.

We used multivariable models to compare those currently receiving HUD housing as-

sistance (public housing, housing choice vouchers, and multifamily housing) with those

who will receive housing assistance within 2 years (the average duration of HUD waitlists)

to account for selection into HUD assistance.

Results. We found reduced odds of fair or poor health for current public housing (odds

ratio [OR] = 0.77; 95% confidence interval [CI] = 0.57, 0.97) and multifamily housing

(OR = 0.75; 95% CI = 0.60, 0.95) residents compared with future residents. Public housing

residents also had reduced odds of psychological distress (OR = 0.59; 95% CI = 0.40, 0.86).

These differences were not mediated by neighborhood-level characteristics, and we did

not find any health benefits for current housing choice voucher recipients.

Conclusions. Housing assistance is associated with improved health and psy-

chological well-being for individuals entering public housing and multifamily

housing programs. (Am J Public Health. 2017;107:571–578. doi:10.2105/AJPH.

2016.303649)

See also Galea and Vaughan, p. 500.

As of 2015, the US Department ofHousing and Urban Development (HUD) provides housing assistance to ap- proximately 5 million families (including nearly 10 million individuals).1 Despite the large public investment in housing assistance, there have been few studies assessing the association between receipt of housing assistanceandhealthstatus.Housingisaprimary dimension of the health-related social context in individuals’ lives.2 Housing quality and lo- cation have been shown to be related to physical and mental health,3,4 exposure to environmentaltoxins,5 andaccesstohealthcare and services.6,7 Studies indicate that improve- ment in housing quality can have beneficial impacts on the health and well-being of low-income families.8

Better housing costs more, and unafford- able housing strains family resources needed to sustain housing stability and

health-promoting investments.9 Especially for low-income families, housing cost burdens increase the risk of eviction and homelessness, which can increase stress and be severely detrimental to mental health.10

Housing assistance provides better and more stable and affordable housing than low-income families otherwise could obtain and may create social contexts conducive to the development of healthy lives.11,12

HUD currently administers several hous- ing assistance programs that are distinguished by their form and assistance structure, the largest of which are public housing, housing choice vouchers, and multifamily housing. Public housing developments are owned and managed by a public housing authority for low-income family occupancy. Public housing tends to result in higher density housing than do other programs, but the demolition of high-poverty tower housing developments in the 1990s was part of a national push to reduce reliance on public housing for housing assistance and to integrate public housing into surrounding communities.13

Housing choice vouchers provide direct housing subsidies that allow recipients to enter the private housing market. Housing choice vouchers are intended to give low- income families the greatest amount of freedom possible in finding a suitable housing unit, in terms of both housing type and neighborhood location.14 Multifamily housing programs involve private housing developments that reserve a certain number of housing units for rent at below-market rates, with the difference in price subsidized by HUD. Compared with public housing, both vouchers and multifamily housing make it more likely that assisted families live in mixed-income developments, although in many cases voucher-assisted families live

ABOUT THE AUTHORS Andrew Fenelon is with the Department of Health Services Administration, University of Maryland, College Park. Patrick Mayne is with the Department of Sociology, Brown University, Providence, RI. Alan E. Simon is with the Office of the Assistant Secretary for Health, US Department of Health and Human Services, Washington, DC. Lauren M. Rossen and Patricia Lloyd are with the National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD. Veronica Helms, Jon Sperling, and Barry L. Steffen are with the Office of Policy Development and Research, US Department of Housing and Urban Development, Washington, DC.

Correspondence should be sent to Andrew Fenelon, 3310B School of Public Health Building, 4200 Valley Drive, College Park, MD 20742 (e-mail: [email protected]). Reprints can be ordered at http://www.ajph.org by clicking the “Reprints” link.

This article was accepted December 22, 2016. Note. The views expressed in this article are those of the authors and do not necessarily reflect the official position of the US Centers for

Disease Control and Prevention. doi: 10.2105/AJPH.2016.303649

April 2017, Vol 107, No. 4 AJPH Fenelon et al. Peer Reviewed Research 571

AJPH RESEARCH

in similarly disadvantaged neighborhoods before and after receiving vouchers. Multifamily housing developments are often recent construction or building rehabilitation and are more likely to be in urban areas.

The HUD-funded Moving to Oppor- tunity project used an experimental ap- proach to examine the effects of movement out of substandard housing developments and into lower-poverty neighborhoods. Although most Moving to Opportunity evaluation research has found effects to be highly heterogeneous,15 improved housing and neighborhoods appear to be a valid pathway to better health outcomes for in- dividuals, including reductions in obesity, diabetes, and mental health difficulties among adults.16,17

However, the primary goal of the Moving to Opportunity experiment was to examine neighborhood effects rather than the effect of housing assistance. Studies considering the effect of receiving housing assistance versus not receiving assistance on health and economic outcomes have often relied on data from individual public housing author- ities or smaller samples.12 Some studies focusing on particular cities have reported that improvements in housing stability and af- fordability can lead to better educational out- comesforchildren,18,19 butit isunclearwhether this is driven by methodology or particular samples.20 Evidence for the impacts of housing assistanceon adult healthoutcomesissomewhat mixed,21 and there have been no national studies examining physical and mental health benefits for adults gaining access to housing assistance.

We examined whether receiving housing assistance is associated with improved health and well-being using a nationally representative sample of the US population. Specifically, we examined whether entry into housing assistance was associated with better reported health or reduced psychological distress relative to awaiting admission and whether there were differential effects associated with the 3 primary program categories: public housing, housing choice vouchers, and multifamily housing. Furthermore, we explored whether the health effects of housing assis- tance are mediated by neighborhood characteristics.

METHODS We linked the annual National Health

Interview Survey (NHIS) with HUD ad- ministrative data to examine relationships between the housing and health of the HUD housing-assisted population.22 This linkage provided information on the housing assistance history of NHIS respondents during the period of HUD administrative records. The HUD administrative records cover the period 1999 to 2014 and are linked to NHIS survey respondents during the period 1999 to 2012.

Sample To be linkage eligible, NHIS survey re-

spondents had to provide sufficient personally identifiable information: Social Security number (including last 4 digits), date of birth, and gender. The final NHIS–HUD link included 264 031 eligible adult (aged ‡ 18 y) respondents interviewed by NHIS (1999– 2012). At the time of their interview, 8695 individuals (3.4% of the sample) were current housing assistance recipients: 2065 in public housing, 3822 in housing choice vouchers, and 2808 in multifamily housing. Additionally, there were respondents who would enter housing assistance within 2 years of their NHIS interview: 1105 in public housing, 1792 in housing choice vouchers, and 1194 in multifamily housing.

In our analysis of psychological distress, we used the NHIS Sample Adult file (a subset of the total person sample). In this analysis 8049 individuals were currently receiving housing assistance: 1912 in public housing, 3509 in housing choice vouchers, and 2628 in multifamily housing. Additionally, 3362 respondents would enter housing assistance within 2 years: 924 in public housing, 1433 in housing choice vouchers, and 1005 in multifamily housing.

Dependent Variables We assessed the association between health

and housing assistance using reported health status and psychological distress as reported in the interview. Household refer- ence persons rated the health of respondents as excellent, very good, good, fair, or poor, and we dichotomized this response as fair or poor versus good, very good, or excellent.

Reported health status takes into account current health conditions as well as individual health trajectories over time, and it is one of the best known predictors of subsequent mortality.23

Psychological distress, an indicator of mental health, was measured using the Kessler-6 scale,24 which asks respondents how often they experience various feelings of distress. Responses to individual items range from 0 (never) to 4 (almost all the time). The Kessler-6 score is the sum of the values for the 6 items; we recoded this score with 13 or greater denoting “serious psychological distress.” This scale predicts major depressive disorder and best reflects nonspecific psy- chological distress in population-based studies.24

Housing Assistance Status The primary exposure was respondents’

receipt of HUD housing assistance, including the timing and program category of each episode of housing assistance. Previous re- search has been limited by the inability to isolate the effect of housing assistance itself from factors that affect both participation in HUD housing assistance and health out- comes.25,26 To overcome this methodologi- cal limitation, we exploited the timing of assistance, comparing current recipients with those who were not receiving assistance but would enter assisted housing within 2 years (future assistance). As waits for public housing agency–administered housing assistance average 2 years,1 these future assistance individuals were likely to resemble those on HUD waitlists. These individuals thus formed an excellent comparison group for those currently receiving assistance.27 This approach allowed us to account for un- observed differences between individuals who obtained housing assistance and those who did not (e.g., selection bias).

We categorized respondents by timing and program category, with each respondent appearing in only 1 group for each variable. There may be some selection in which re- spondents enter a particular housing program, reflecting either self-selection or geographic differences in the number of housing units associated with each housing program; therefore, we coded respondents in the future assistance category on the basis of

AJPH RESEARCH

572 Research Peer Reviewed Fenelon et al. AJPH April 2017, Vol 107, No. 4

the program category corresponding with the first HUD program they entered (e.g., future public housing).

The waitlist times of those in the future assistance group varied across local public housing agencies, and we performed a sup- plemental analysis (Appendix A, available in a supplement to the online version of this article at http://www.ajph.org) using only those for whom waitlist information was available in the data linkage. The waitlist data were limited in that information on waitlist entry dates may have been of questionable quality in the HUD data and were unavailable for multifamily housing.22 However, the similarity of these results to our main results suggests that the future assistance group represented a good approximation of those on HUD waitlists. The models in Appendix B (available as a supplement to the online version of this article at http://www.ajph. org) consider each housing program category separately.

Covariates We adjusted for individual characteristics,

including age (linear and quadratic terms), gender, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, other), family size, presence of children in the household, education (< high school, high school, > high school), family income to poverty ratio (< 50%, 50%–99%, 100%–199%, ‡ 200%), employment status (employed, unemployed, not in labor force), insurance status (private insurance, public insurance, uninsured, unknown), other assistance program partici- pation (the Supplemental Nutrition Assis- tance Program and the Special Supplemental Nutrition Program for Women, Infants, and Children), and year of interview. Models including year as a categorical vari- able showed no differences from the main analysis.

We also included socioeconomic and demographic characteristics of census tracts, because receiving housing assistance may be associated with changes in neighborhood disadvantage for recipients. We obtained census tract characteristics from the 2000 Census Summary File 3 and the 2005 to 2009 American Community Survey and used the National Center for Health Statistics– linked geocode files. Characteristics included

median income, poverty rate, percentage unemployed, percentage receiving public assistance, percentage in professional occu- pations, percentage female-headed house- holds, percentage renter-occupied housing, and percentage living in a different house 5 years before.

We used principal components analysis to isolate combinations of neighborhood factors. This produced 2 components, which to- gether accounted for 69% of the total variance of these variables. The first component (neighborhood disadvantage) was loaded with the poverty rate (0.42) and the per- centage receiving public assistance (0.38). The second component (neighborhood in- stability) was associated with percentage living in a different house 5 years before (0.47) and percentage renter-occupied housing (0.40). We also included a measure of neighborhood racial composition, catego- rizing tracts as mostly White (> 90%), mostly Black (> 90%), mostly Hispanic (> 50%), or mixed race (all other compositions).

Analysis For each health outcome, we used logistic

regression to examine the relationship be- tween housing assistance timing (future vs current assistance) and reported fair or poor health or psychological distress. We included an interaction between program category and timing to test for program category differences in the effect of moving from future assistance to current assistance. We calculated predictive margins to obtain esti- mated probabilities of each health outcome by housing assistance timing and HUD program category using the margins com- mand in Stata version 13 (StataCorp LP, College Station, TX). This technique has been shown to improve some of the un- certainties in the interpretation of interaction effects in logistic regression models.28 Our reported probabilities are average marginal effects across all covariate distributions, but we found no major differences when calculating conditional marginal effects for different race, income, and education levels (results not shown).

Our analyses account for the complex survey design of the NHIS and incorporate weights created by the National Center for Health Statistics that account for both

linkage eligibility and nonresponse to make estimates representative of the civilian noninstitutionalized US population.22 To consider the possibility that the effects of housing assistance on health outcomes reflect the duration of time spent in assisted housing, our analysis shown in Appendix C (available as a supplement to the online version of this article at http://www.ajph. org) restricted current recipients to those who had been receiving housing assistance for less than 2 years. Appendix D (available as a supplement to the online version of this article at http://www.ajph.org) presents the average marginal effects separately by race/ ethnicity.

RESULTS Socioeconomic and demographic char-

acteristics of the analytic sample by timing of housing assistance and program category are shown in Table 1. Overall, HUD housing assistance recipients had low levels of socioeconomic status. More than 50% of all individuals in the sample had family income below the federal poverty level, in- cluding 65% of current recipients in each program. Current and future recipients had similar demographic profiles, with no differences in terms of gender, race/ethnicity, or education. When compared with future recipients, current residents were older, had lower family income, were less likely to be employed, and were more likely to have public health insurance, which may reflect the wait for housing assistance. Recipients of housing choice vouchers tended to ex- perience lower neighborhood disadvantage than did those in other program categories. Current public housing residents were less likely to report serious psychological distress than were future residents (P < .05).

The models depicted in Table 2 predict the odds of fair or poor health and serious psychological distress status as a function of housing assistance timing and program cate- gory. These models include an interaction between timing of housing assistance and program category that enabled us to exam- ine differences in the effect of housing assistance among different program cate- gories. The first column shows reported fair or poor health. The main effect of housing

AJPH RESEARCH

April 2017, Vol 107, No. 4 AJPH Fenelon et al. Peer Reviewed Research 573

timing in this model refers specifically to housing choice vouchers, the omitted cate- gory. There were no differences by housing assistance timing for those who had housing

choice vouchers (odds ratio [OR] = 1.05; 95% confidence interval [CI] = 0.90, 1.25). We obtained the housing timing coefficients for public housing and multifamily housing

by multiplying the main effect by the in- teraction. Current residents had lower odds of fair or poor health than did future public housing residents (1.05 · 0.73 = 0.77; 95%

TABLE 1—Descriptive Characteristics by Housing Assistance Timing and Program Category: Linked NHIS–HUD Data, United States, 1999–2012

Public Housing Housing Choice Vouchers Multifamily Housing

Characteristic Currenta Futureb Currenta Futureb Currenta Futureb

No. 2065 1105 3822 1792 2808 1194

HUD-assisted, y, mean 6SD 3.1 60.5 . . . 3.1 60.4 . . . 3.1 60.5 . . .

Age, y, mean 6SD 49.1 61.2** 43.4 61.0 42.5 60.4** 39.2 60.5 52.5 61.3** 47.5 61.0

Women, % (SE) 72.3 (1.5) 71.1 (1.7) 78.8 (0.9) 76.4 (1.3) 73.4 (1.4) 70.4 (1.6)

Household size, mean 6SD 2.8 61.6 2.9 61.8 3.0 61.5 3.0 61.7 2.9 61.6 2.7 61.6

Children present, % (SE) 40.9 (2.7) 47.4 (2.5) 55.4 (1.1)* 61.6 (1.5) 33.2 (2.6) 37.5 (2.1)

Race/ethnicity, % (SE)

Non-Hispanic White 34.0 (3.2) 34.0 (2.4) 37.5 (1.4) 37.7 (1.6) 50.3 (2.8) 55.9 (2.4)

Non-Hispanic Black 39.3 (3.3) 42.1 (2.5) 37.9 (1.2) 38.5 (1.6) 27.9 (2.4) 26.3 (2.0)

Non-Hispanic other 5.6 (1.9) 3.8 (0.8) 4.9 (0.6) 4.4 (0.7) 5.4 (0.8) 4.3 (0.8)

Hispanic 21.1 (2.3) 20.1 (2.3) 19.7 (0.9) 19.4 (1.3) 16.4 (1.6) 13.5 (1.7)

Education, % (SE)

< high school 40.8 (2.0) 39.3 (1.7) 32.5 (0.9) 32.6 (1.3) 38.7 (1.5) 36.7 (1.7) High school 29.9 (1.4) 33.7 (1.6) 29.8 (0.9) 28.5 (1.3) 31.7 (1.3) 32.0 (1.7)

> high school 29.2 (1.4) 27.0 (1.7) 37.7 (1.0) 38.8 (1.6) 29.5 (1.3) 31.3 (1.6)

Poverty status, % of FPL (SE) ** ** **

< 50 23.8 (1.7) 19.9 (1.5) 23.4 (0.8) 22.6 (1.2) 22.3 (1.7) 17.4 (1.4) 50–99 41.4 (1.6) 34.9 (1.7) 42.5 (0.9) 35.4 (1.4) 42.0 (1.2) 29.5 (1.2)

100–199 26.5 (1.3) 30.3 (1.7) 28.1 (0.9) 29.7 (1.4) 29.9 (1.3) 37.9 (1.7)

‡ 200 8.3 (0.7) 14.9 (1.5) 6.0 (0.5) 12.2 (1.0) 5.8 (0.6) 15.1 (1.3)

Employment status, % (SE) * * *

Employed 28.9 (1.5) 36.2 (1.8) 35.4 (0.9) 39.9 (1.5) 22.7 (1.4) 31.9 (1.9)

Unemployed 8.5 (0.9) 8.1 (1.0) 11.5 (0.7) 10.6 (0.7) 8.5 (1.0) 9.3 (1.2)

Not in labor force 61.2 (1.9) 54.6 (1.9) 51.1 (1.0) 47.9 (1.5) 66.5 (2.0) 56.2 (1.9)

Missing 1.4 (0.4) 1.1 (0.4) 1.9 (0.2) 1.6 (0.4) 2.3 (0.4) 2.5 (0.5)

Insurance status, % (SE) ** ** **

Private insurance 9.0 (0.9) 17.7 (1.5) 11.4 (0.6) 13.3 (1.1) 5.7 (0.6) 14.5 (1.3)

Public insurance 73.1 (1.5) 62.5 (2.0) 70.0 (1.0) 63.8 (1.5) 80.5 (1.3) 65.6 (1.9)

No insurance 17.7 (1.3) 19.4 (1.6) 18.4 (0.8) 22.9 (1.3) 13.5 (1.2) 20.0 (1.6)

Received SNAP or WIC, % (SE) 56.8 (2.2)** 45.9 (2.0) 61.3 (1.0)** 53.9 (1.5) 52.3 (2.0)** 41.9 (2.1)

Health status,c % (SE)

Fair or poor 35.2 (1.3) 37.6 (1.8) 36.0 (1.0) 34.0 (1.3) 32.1 (1.2) 34.4 (1.7)

Good, very good, or excellent 64.8 (1.3) 62.4 (1.8) 64.0 (1.0) 66.0 (1.3) 67.9 (1.2) 65.6 (1.7)

Psychological distress,c % (SE) *

Serious distress 10.1 (0.9) 13.8 (1.5) 12.8 (0.8) 11.7 (1.0) 12.2 (0.9) 11.1 (1.2)

Not serious distress 89.9 (0.9) 86.2 (1.5) 87.2 (0.8) 88.3 (1.0) 87.8 (0.9) 88.9 (1.2)

Census tract characteristics, % (SE)

Neighborhood disadvantage 1.4 (0.1) 1.1 (0.1) 0.5 (0.0)** 0.7 (0.0) 0.8 (0.1)** 0.5 (0.1)

Neighborhood instability –0.2 (0.1)* 0.1 (0.1) –0.1 (0.0)* 0.0 (0.0) –0.1 (0.1) –0.1 (0.0)

Continued

AJPH RESEARCH

574 Research Peer Reviewed Fenelon et al. AJPH April 2017, Vol 107, No. 4

CI = 0.58, 0.99). Current multifamily hous- ing residents also had lower odds of fair or poor health than did future residents (OR = 0.77; 95% CI = 0.60, 0.98).

Using the model shown in the second column, we examined the effects of housing assistance on the odds of reporting serious psychological distress. There was a signifi- cant reduction in the odds of psychological distress for residents of public housing. Current residents had lower odds of serious psychological distress than did future public housing residents (OR = 0.59; 95% CI = 0.40, 0.88). We did not find a difference in psychological distress between current and future recipients of housing choice vouchers or residents of multifamily housing.

The marginal probabilities (Table 3) in- dicate that the percentage reporting fair or poor health was reduced by 4.8 percentage points (95% CI = 0.8, 8.9; P = .02) for current public housing residents and by 4.8 per- centage points (95% CI = 1.13, 8.5; P = .01) for current multifamily housing residents compared with future residents. Current public housing residents also experienced a reduction in serious psychological distress of 5.4 percentage points compared with future residents (95% CI = 1.8, 9.1; P = .002), from 14.7% to 9.3%.

DISCUSSION To our knowledge, this is the first

nationally representative study of the

relationship between HUD housing assis- tance programs and adult health outcomes in the United States. It is also the first, to our knowledge, to explore the heterogeneous effects of major housing assistance programs on health. Our results demonstrate that receiving housing assistance is associated with a benefit for low-income adults but that this effect depends on the specific program category of assistance. Entering public housing and multifamily housing led to an improvement in reported health status rela- tive to not yet receiving these types of assistance. The likelihood of serious psycho- logical distress was also lower among in- dividuals residing in public housing than among individuals who had yet to enter public housing. We did not find consistent health benefits for those who received housing choice vouchers, which constituted almost half of all HUD-assisted units.1

Although the total number of housing units managed under all HUD programs has in- creased by 20% since 1997, public housing units as a proportion of all units have de- creased from 32% in 1997 to 22% in 2015.1

Adults who live in public housing are considerably less likely to experience serious psychological distress than are future public housing residents, an effect not observed for other housing assistance programs. Although this is a notable difference, future public housing residents have poorer health than do those who enter housing choice vouchers or multifamily housing. The improvement may reflect the presence of

a network effect in which public housing reinforces social networks in neighborhoods with particularly high disadvantages. Historically, public housing projects were exceptionally dense, and the density of public housing today remains higher than the av- erage in US cities.13 In contrast with housing choice voucher programs, public housing may consolidate limited social resources for low-income families.29 A concentration of individuals with similar social and eco- nomic experiences, health needs, and requirements for services may provide an informal source of information and support.30

Indeed, although public housing may increase neighborhood disadvantage, it does not appear to reduce social capital for low-income residents.31 Future research should work to identify the particular mechanisms that link public housing residence to improved health.

Our results are unable to demonstrate consistent health benefits for recipients of housing choice vouchers. This may partially reflect heterogeneity in the effect of vouchers across subgroups,15 which may make it difficult to identify effects. Effect heterogeneity may stem from programmatic changes to the voucher program that oc- curred between 2005 and 2013, including changes to the rent ceiling, changes in the size of voucher subsidies across public housing agencies, and the localization of fair market rents for voucher units.32 Collinson and Ganong33 demonstrate that whereas increasing the subsidy to voucher households

TABLE 1—Continued

Public Housing Housing Choice Vouchers Multifamily Housing

Characteristic Currenta Futureb Currenta Futureb Currenta Futureb

Racial/ethnic composition, % (SE)

> 90 White 13.2 (3.3) 16.2 (2.0) 20.0 (1.4) 19.2 (1.5) 23.3 (3.3) 26.8 (2.4) > 50 Hispanic 14.9 (2.8) 15.8 (2.4) 11.0 (0.8) 11.9 (1.0) 7.9 (1.4) 5.8 (1.1) > 90 Black 6.7 (1.6) 7.4 (1.6) 5.1 (0.6) 5.5 (0.7) 5.3 (1.6) 4.4 (0.8) Mixed race 65.2 (4.1) 60.7 (3.0) 64.0 (1.6) 63.4 (1.7) 63.7 (3.6) 63.1 (2.5)

Note. FPL = federal poverty level; HUD = US Department of Housing and Urban Development; NHIS = National Health Interview Survey; SNAP = Supplemental Nutrition Assistance Program; WIC = Special Supplemental Nutrition Program for Women, Infants, and Children. Statistical differences in distribution are between current and future in each program category. For categorical variables, significance refers to the c2 test. FPL determined by the US Census Bureau in the year of interview. aReceiving housing assistance at interview. bTo receive housing assistance within 2 years of interview. cTotals are age-standardized.

*P < .05; **P < .01.

AJPH RESEARCH

April 2017, Vol 107, No. 4 AJPH Fenelon et al. Peer Reviewed Research 575

TABLE 2—Odds Ratios (95% Confidence Intervals) of Fair or Poor Health and Serious Psychological Distress as a Function of Housing Timing and Program Category: Linked NHIS–HUD Data, United States, 1999–2012

Housing Timing (Future)a Fair or Poor Health (n = 12 786), OR (95% CI) Serious Psychological Distress (n = 11 411), OR (95% CI)

Current 1.06 (0.90, 1.25) 1.05 (0.83, 1.33)

Program category

Housing choice vouchers (Ref) 1 1

Public housing 1.29 (1.04, 1.61) 1.38 (1.00, 1.90)

Multifamily housing 1.06 (0.86, 1.31) 0.92 (0.68, 1.34)

Interactions

Current · public housing 0.73 (0.56, 0.95) 0.56 (0.46, 0.83) Current · multifamily housing 0.73 (0.57, 0.93) 0.95 (0.64, 1.41)

Race/ethnicity

Non-Hispanic White (Ref) 1 1

Non-Hispanic Black 0.92 (0.86, 1.09) 0.56 (0.46, 0.68)

Non-Hispanic other 0.92 (0.70, 1.21) 0.87 (0.58, 1.29)

Hispanic 1.06 (0.89, 1.27) 1.07 (0.84, 1.35)

Household size 0.99 (0.92, 1.08) 1.02 (0.83, 1.30)

Children in household 0.71 (0.59, 0.87) 0.60 (0.42, 0.87)

Education

Less than high school (Ref) 1 1

High school 0.82 (0.73, 0.92) 0.84 (0.69, 1.03)

More than high school 0.74 (0.66, 0.83) 0.88 (0.72, 1.08)

Employment status

Employed (Ref) 1 1

Unemployed 1.45 (1.16, 1.81) 2.08 (1.44, 2.99)

Not in labor force 2.98 (2.59, 3.44) 2.54 (2.18, 3.35)

Poverty status, % of FPL

< 50 (Ref) 1 1 50–99 1.01 (0.88, 1.16) 0.83 (0.68, 1.03)

100–199 0.96 (0.20, 1.11) 0.71 (0.48, 1.02)

‡ 200 0.79 (0.61, 0.97) 0.61 (0.41, 0.93)

Insurance status

Private (Ref) 1 1

Public insurance 1.67 (1.33, 2.10) 1.30 (0.89, 1.90)

No insurance 1.35 (1.06, 1.72) 1.18 (0.79, 1.76)

Other assistance

Received SNAP 1.28 (1.12, 1.45) 1.33 (1.10, 1.62)

Received WIC 0.65 (0.50, 0.84) 1.08 (0.76, 1.52)

Census tract characteristics

Neighborhood disadvantage 1.00 (0.96, 1.05) 1.02 (0.94, 1.11)

Neighborhood instability 1.00 (0.95, 1.06) 0.94 (0.87, 1.02)

Racial composition

> 90% White (Ref) 1 1 > 90% Black 0.99 (0.75, 1.30) 0.94 (0.57, 1.54) > 50% Hispanic 0.89 (0.70, 1.12) 0.74 (0.52, 1.06) Mixed race 1.00 (0.87, 1.15) 0.85 (0.67, 1.06)

Note. CI = confidence interval; FPL = federal poverty level; HUD = US Department of Housing and Urban Development; NHIS = National Health Interview Survey; OR = odds ratio; SNAP = Supplemental Nutrition Assistance Program; WIC = Special Supplemental Nutrition Program for Women, Infants, and Children. Models included age, age2, gender, interview year, individual-level economic and demographic characteristics, and neighborhood-level characteristics. FPL determined by the agency in the year of interview. aNoncurrent but will receive housing assistance within 24 months.

AJPH RESEARCH

576 Research Peer Reviewed Fenelon et al. AJPH April 2017, Vol 107, No. 4

across a metropolitan area had little impact on neighborhood quality outcomes, policies that relate subsidy increases to area rents had positive impacts on neighborhood outcomes.

Consequently, we may be unable to identify health benefits of housing choice vouchers if these policy changes are directed to particularly high-need communities. Although having an increase in disposable resources as a result of having housing assis- tance may contribute to the health benefits we observed for public housing and multi- family housing, the policy changes may reduce the salience of income effects for housing choice vouchers.14 Furthermore, effects estimated separately by race/ethnicity (Appendix D) suggest that programs may have differential benefits for population subgroups, which should be an area for future research.

Although we are unable to examine the effect of changing neighborhoods over time, the effects of housing assistance on health do not appear to be mediated by neighbor- hood characteristics. We do find that housing choice vouchers provide families access to less disadvantaged neighborhoods than do public housing and multifamily housing. This is consistent with evidence that vouchers lead to reduced neighborhood disadvantage,34 but we did not find that this leads to improved health and well-being for voucher recipients. This may imply that the health benefits of improved neighborhoods may develop over a longer period of time.17 Future work should establish the precise relationship between

housing program, neighborhood character- istics, and health outcomes.

Limitations and Strengths Our analysis has limitations that should be

noted. First, the NHIS–HUD link contained longitudinal information on housing status but did not capture changes in health over time as a function of housing assistance. Therefore, we were unable to explicitly model the relationship between duration of housing assistance and health. Our results in Appendix C suggest that the observed health benefits of housing assistance are not driven by duration effects, because adults who have lived in HUD housing for less than 2 years show similar benefits. Because we compared individuals living in HUD-assisted housing to those who would enter within 2 years, health differences that manifest relatively quickly were more easily captured.

Second, our analytical strategy compared those receiving housing assistance with those entering assisted housing within 2 years. Al- though this helps to adjust for unobserved characteristics of the assisted housing pop- ulation, it means that we can only generalize our results to the population who receives assistance, and not necessarily to the broader population who might have been eligible for assistance but did not receive it.

Third, our analysis did not explicitly compare housing assistance recipients with individuals who were actually known to be on HUD waitlists. Waitlist information is

unavailable for multifamily housing programs in HUD administrative records, but our re- sults shown in Appendix A are nearly identical for public housing and housing choice vouchers when restricting future recipients to those determined to be on a waitlist at the time of interview.

Finally, we were unable to examine the effects of providing housing to formerly homeless individuals, because NHIS is a household survey. However, considering the effect of homelessness and health, the inclusion of the formerly homeless might expand the benefits we observed for housing assistance.

The strengths of this study are the use of a large national sample from linked survey and administrative data, evaluating effects of specific program categories, and consid- ering the role of neighborhood-level characteristics. Additionally, our analytic technique, comparing current housing assis- tance recipients with future recipients, re- duced issues of selection bias in the receipt of assistance. Because the data linkage pro- cedure confirmed that future recipients would enter HUD-assisted housing within 2 years of the interview, we implicitly accounted for differences between groups in motivation to participate and the local availability of HUD-assisted housing, which may be problematic using other study designs.

Public Health Implications We have provided evidence that HUD

public housing is associated with a reduction

TABLE 3—Probabilities of Fair or Poor Health and Serious Psychological Distress by Housing Assistance Timing and Program Category: Linked NHIS–HUD Data, United States, 1999–2012

Timing Public Housing, % (95% CI) Housing Choice Vouchers, % (95% CI) Multifamily Housing, % (95% CI)

Fair or poor health

Currenta 34.8 (32.6, 37.0) 35.8 (34.1, 37.5) 31.2 (29.5, 33.0)

Futureb 39.6 (36.2, 42.9) 34.8 (32.3, 37.3) 36.0 (32.9, 39.0)

Difference 4.8 (0.8, 8.9) –1.0 (–4.0, 1.9) 4.8 (1.1, 8.5)

Serious psychological distress

Current 9.3 (7.7, 10.9) 12.7 (11.2, 14.0) 11.4 (9.8, 13.0)

Future 14.7 (11.6, 17.6) 12.1 (10.2, 14.0) 11.4 (8.7, 14.0)

Difference 5.4 (1.8, 9.1) –0.6 (–2.9, 1.8) 0.0 (–3.1, 3.2)

Note. CI = confidence interval; HUD = US Department of Housing and Urban Development; NHIS = National Health Interview Survey. Predicted probabilities are average marginal probabilities, estimated as the average across all covariate levels. Although presented differences may differ slightly from those estimated at the average of all the covariates, the substantive interpretation is unchanged. aReceiving housing assistance at interview. bTo receive housing assistance within 2 years of interview.

AJPH RESEARCH

April 2017, Vol 107, No. 4 AJPH Fenelon et al. Peer Reviewed Research 577

in the likelihood of poor health and psy- chological distress. Because of recent interest in approaches to improve health through interventions that target community, social– structural, and environmental factors,35 un- derstanding the relationship between HUD housing assistance and health is likely to be of interest and value to researchers and policymakers.

CONTRIBUTORS A. Fenelon initiated the study, designed the analytical approach, wrote the first draft of the article, and supervised the project. A. Fenelon, A. E. Simon, L. M. Rossen, V. Helms, and P. Lloyd participated in statistical data analysis. P. Mayne, A. E. Simon, and L. M. Rossen participated in the study design and analytical approach. V. Helms, P. Lloyd, J. Sperling, and B. L. Steffen participated in the acquisition of data and technical support. All authors participated in the interpretation of data and critical revision of the article.

ACKNOWLEDGMENTS We acknowledge technical and data support from the Special Projects Branch, Office of Analysis and Epide- miology,NationalCenterforHealthStatisticsandfromthe Office of Policy Development and Research, US De- partment of Housing and Urban Development. We are grateful to Lisa Mirel, Luisa Franzini, and Michel Boudreaux for helpful comments on earlier versions of the article.

HUMAN PARTICIPANT PROTECTION No protocol approval was necessary because no human participants were involved in this study.

REFERENCES 1. US Department of Housing and Urban Development. Picture of subsidized households. 2015. Available at: https://www.huduser.gov/portal/datasets/picture/ yearlydata.html. Accessed March 3, 2016.

2. Diez Roux AV, Mair C. Neighborhoods and health. Ann N Y Acad Sci. 2010;1186(1):125–145.

3. Burgard SA, Seefeldt KS, Zelner S. Housing instability and health: findings from the Michigan Recession and Recovery Study. Soc Sci Med. 2012;75(12):2215–2224.

4. Evans GW, Wells NM, Moch A. Housing and mental health: a review of the evidence and a methodological and conceptual critique. J Soc Issues. 2003;59(3):475–500.

5. Adamkiewicz G, Zota AR, Fabian MP, et al. Moving environmental justice indoors: understanding structural influences on residential exposure patterns in low-income communities. Am J Public Health. 2011;101(suppl 1): S238–S245.

6. Reid KW, Vittinghoff E, Kushel MB. Association between the level of housing instability, economic standing and health care access: a meta-regression. J Health Care Poor Underserved. 2008;19(4):1212–1228.

7. Hwang SW, Gogosis E, Chambers C, Dunn JR, Hoch JS, Aubry T. Health status, quality of life, residential stability, substance use, and health care utilization among adults applying to a supportive housing program. J Urban Health. 2011;88(6):1076–1090.

8. Cutts DB, Meyers AF, Black MM, et al. US housing insecurity and the health of very young children. Am J Public Health. 2011;101(8):1508–1514.

9. Pollack CE, Griffin BA, Lynch J. Housing affordability and health among homeowners and renters. Am J Prev Med. 2010;39(6):515–521.

10. Tsai AC. Home foreclosure, health, and mental health: a systematic review of individual, aggregate, and contextual associations. PLoS One. 2015;10(4): e0123182.

11. Rauh VA, Landrigan PJ, Claudio L. Housing and health. Ann N Y Acad Sci. 2008;1136(1):276–288.

12. Jacob BA, Kapustin M, Ludwig J. The impact of housing assistance on child outcomes: evidence from a randomized housing lottery. Q J Econ. 2015;130(1): 465–506.

13. Schmoke KL, Cisneros HG, Engdahl L. From Despair to Hope: HOPE VI and the New Promise of Public Housing in America’s Cities. Washington, DC: Brookings Institution Press; 2010.

14. US Department of Housing and Urban Development. Office of Housing Choice Vouchers. 2015. Available at: https://portal.hud.gov/hudportal/HUD?src=/program_ offices/public_indian_housing/programs/hcv. Accessed September 15, 2015.

15. Nguyen QC, Rehkopf DH, Schmidt NM, Osypuk TL. Heterogeneous effects of housing vouchers on the mental health of US adolescents. Am J Public Health. 2016; 106(4):755–762.

16. Ludwig J, Duncan GJ, Gennetian LA, et al. Long-term neighborhood effects on low-income families: evidence from Moving to Opportunity. Am Econ Rev. 2013;103(3): 226–231.

17. Leventhal T, Brooks-Gunn J. Moving to opportunity: an experimental study of neighborhood effects on mental health. Am J Public Health. 2003;93(9):1576–1582.

18. Cunningham M, MacDonald G. Housing as a Platform for Improving Education Outcomes Among Low-Income Children. Washington, DC: Urban Institute; 2012.

19. Newman SJ, Holupka CS. Housing affordability and child well-being. Hous Policy Debate. 2015;25(1): 116–151.

20. Jacob BA, Ludwig J. The effects of housing assistance on labor supply: evidence from a voucher lottery. Am Econ Rev. 2012;102(1):272–304.

21. Newman SJ. Does housing matter for poor families? A critical summary of research and issues still to be re- solved. J Policy Anal Manage. 2008;27(4):895–925.

22. Lloyd PC, Helms VE. NCHS–HUD Linked Data: Analytic Considerations and Guidelines. Hyattsville, MD: Office of Analysis and Epidemiology, National Center for Health Statistics; 2016.

23. Jylhä M. What is self-rated health and why does it predict mortality? Towards a unified conceptual model. Soc Sci Med. 2009;69(3):307–316.

24. Kessler RC, Andrews G, Colpe LJ, et al. Short screening scales to monitor population prevalences and trends in non-specific psychological distress. Psychol Med. 2002;32(6):959–976.

25. Fertig AR, Reingold DA. Public housing, health, and health behaviors: is there a connection? J Policy Anal Manage. 2007;26(4):831–859.

26. Digenis-Bury EC, Brooks DR, Chen L, Ostrem M, Horsburgh CR. Use of a population-based survey to describe the health of Boston public housing residents. Am J Public Health. 2008;98(1):85–91.

27. Baicker K, Taubman SL, Allen HL, et al. The Oregon experiment—effects of Medicaid on clinical outcomes. N Engl J Med. 2013;368(18):1713–1722.

28. Ai C, Norton EC. Interaction terms in logit and probit models. Econ Lett. 2003;80(1):123–129.

29. Venkatesh SA. American Project: The Rise and Fall of a Modern Ghetto. Cambridge, MA: Harvard University Press; 2000.

30. Pickett KE, Wilkinson RG. People like us: ethnic group density effects on health. Ethn Health. 2008;13(4): 321–334.

31. Reingold DA, Van Ryzin GG, Ronda M. Does urban public housing diminish the social capital and labor force activity of its tenants? J Policy Anal Manage. 2001; 20(3):485–504.

32. Sard B, Rice D. Realizing the Housing Voucher Program’s Potential to Enable Families to Move to Better Neighborhoods. Washington, DC: Center on Budget and Policy Priorities; 2016.

33. Collinson RA, Ganong P. The incidence of housing voucher generosity. 2015. Available at: http://scholar. harvard.edu/files/ganong/files/collinsonganong105. pdf?m=1446135036. Accessed January 10, 2017.

34. Boston TD. The effects of revitalization on public housing residents: a case study of the Atlanta Housing Authority. J Am Plann Assoc. 2005;71(4):393–407.

35. Krieger J, Higgins DL. Housing and health: time again for public health action. Am J Public Health. 2002;92(5): 758–768.

AJPH RESEARCH

578 Research Peer Reviewed Fenelon et al. AJPH April 2017, Vol 107, No. 4

Copyright of American Journal of Public Health is the property of American Public Health Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.