Association for Public Policy Analysis & Management

Improved housing is an important means to improve health, and housing assistance programs seek to provide high-quality and stable housing for low-income families. Understanding the health needs of the housing assistance population and the effectiveness of housing assistance to improve health requires data that include measures of health and indicators of housing assistance. The most easily accessible data of this sort are health surveys that include questions on housing assistance participation. Previous research suggests that like other public assistance programs, self-reports of participation in housing assistance programs are characterized by a high false-negative rate (true participants who report that they do not participate). However, the literature examining the validity of housing assistance participation has focused on non-health surveys and it provides little evidence on the extent to which misreports vary across demographic and health status variables. In this paper we examine the validity of the housing assistance question in the National Health Interview Survey (NHIS) using a version of the survey that has been linked to administrative data from the Department of Housing and Urban Development (HUD). NHIS respondents from 1999-2012 were linked to the HUD record using social security number, date of birth, and sex. A total of 27,555 NHIS respondents were identified in the HUD record as participating in a housing assistance program during this period.

Because the NHIS question ascertains participation in any housing assistance program (HUD and non-HUD administered) we are constrained to estimating false-negative rates for HUD participation. We estimate the proportion of known HUD participants that are misclassified as having no assistance (the false negative rate) overall and across key subgroups. These subgroups include demographics (age, race, sex, and geographic region), health (self-reported health status and chronic conditions), assistance type (public housing, multi-family housing, housing choice vouchers) and length of program tenure. Finally, we use a waitlist technique to examine the impact of housing assistance on reported health and health care access, comparing those receiving housing assistance to those on the HUD waitlist. This method seeks to adjust for unobserved differences between housing assistance recipients and non-recipients. We compare the association of housing assistance and health outcomes estimated from multivariable regressions before and after partially correcting the survey based housing assistance indicator using the administrative records. Our results will assess the accuracy of the survey self-report of assistance participation as well as whether discrepancies between the self-report and the direct measurement of participation impact statistical estimates of the effect of housing assistance on health.