Rental Housing Assistance and Health: Evidence From the Survey of Income and Program Participation

Abstract

Interest in the health impacts of renter housing assistance has grown in the wake of heated national discussions on health care and social welfare spending. Assistance may improve renters’ health by offering (a) low, fixed housing costs; (b) protection against eviction; and (c) access to better homes and neighborhoods. Using data from the Survey of Income and Program Participation and econometric analysis, I estimate the effect of receiving assistance from the public housing or Section 8 voucher programs on low-income renters’ reported health status and spending. Assisted renters spent less on health care over the year than unassisted low-income renters did, after controlling for other characteristics. This finding suggests that assisted housing leads to health benefits that may reduce low-income renters’ need to purchase health services. Voucher holders’ lower expenditures are influenced by their low, fixed housing costs, but public housing residents’ lower expenditures are not explained by existing theory.

Keywords:

• Rental housing
• health care
• public housing
• low-income housing

Acknowledgments

The author would like to thank Kirk McClure and other participants in the 2016 Association of Collegiate School of Planning conference preorganized session “Outcomes in the Housing Choice Voucher Program: Location and Beyond” for their encouragement and helpful comments on an early version of this research.

Notes

1. About 90–98% of people living in subsidized housing were in a low-income family across the panels. About 62–68% of the total sampled people living in renter households were in a low-income family.

2. This method entails dividing household income by the square root of household size: Adjusted household income = household income/(household size^0.5). This adjustment assumes that a two-person household only needs 1.414 times the income of a one-person household to be equally well off (Burkhauser & Larrimore, 2014).

3. This sample is smaller than the complete sample of original respondents who were low-income renters age 25 to 69 (26,073), because information on participants’ experience of eviction and housing and neighborhood conditions was only gathered once during the 2001 and 2004 panels and not until several years into the panel, when some attrition had occurred. SIPP has relatively high attrition rates. About 20% of the original sample typically drops out during the course of the study period; about half of all attrition occurs between the first and second interviews (Van Hook & Glick, 2007). Information on eviction and housing and neighborhood conditions was gathered twice during the 2008 panel, but similarly not until several years in, after some attrition had occurred. The second set of 2008 observations on participants’ housing and neighborhood conditions was excluded from the analysis to avoid over-weighting that panel’s trends in the analysis.

4. This sample is larger than the first sample, because information on participants’ health and housing costs was collected 3 times (once a year) during the 2001 and 2008 panels and 2 times (once a year) during the 2004 panels; information on receiving housing subsidies and demographic and socioeconomic conditions was collected every 4 months in all panels. Information on changes in participants’ health and housing situation was available twice in the 2001 and 2008 panels and once in the 2004 panel.

5. Spending less money on health care was associated with better health among assisted and unassisted renters. Members of both groups who reported being in good health spent on average about $100 less per year on health care than those who did not report being in good health (difference significant at the 5% level).

6. Most of the control variables were statistically associated with receiving public housing or a Section 8 voucher. Two exceptions were being Asian and working part time, which were categories of the variables race and employment status, respectively. Results from this stage of the propensity score matching analysis are available upon request.

7. The matching method was one-to-one matching with replacement. Abadie–Imbens standard errors are reported. The caliper is 0.02 standard deviations of the propensity score.

8. I used the Rosenbaum bounds method to assess how sensitive the estimated effects of having assisted housing on health spending were to potential hidden bias (see Rosenbaum, 2002). The analysis did not show evidence of overt hidden bias.

9. I tested interaction effects between housing burden and moving into public or Section 8 housing in unreported iterations of the model, but none of these effects was statistically significant at the 10% level or higher.