Modeling the Intersectionality of Lifetime Homelessness In the United States

Background and Purpose: Homelessness is rising at an unprecedented rate in the United States (U.S.). Lifetime prevalence of homelessness varies widely across sociodemographic groups, reflecting broader structural inequities in housing and social welfare systems. Although disparities in homelessness are well documented, its intersectional patterning remains unexplored at the population level. We applied a novel intersectional approach to examine variations in the lifetime prevalence of homelessness among U.S.

Methods: We analyzed 2012–2019 data from the Health and Retirement Survey (HRS), a representative sample of U.S. adults aged 50 and older. Lifetime homelessness was defined by affirmative responses to the questions, “Have you ever been homeless?” or “Before you were age 50, were you ever homeless for 1 month or more?” (depending on survey version). The probability of lifetime homelessness was modeled using logistic Intersectional Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (I-MAIHDA), theoretically oriented multilevel approach for examining intersectional inequities. Sociodemographic predictors included birth cohort (pre-1945, baby boomers), race (White, Black), gender (female, male), education (high school or less, beyond high school), and veteran status (non-veteran, veteran). Following the I-MAIHDA, we fitted a simple intersectional model and an additive main effects model and compared their Variance Partition Coefficients (VPC) to determine the Proportional Change in Variance (PCV). The PCV represents the proportion of between-stratum variance explained by adjustment for sociodemographic characteristics.

Results: A total of 16,149 HRS participants aged 50 and older had valid data on homelessness and sociodemographic characteristics of interest. In the analytic sample, 52% were baby boomers, 20% identified as Black, 41% were male, 51% had education beyond high school, and 17% were veterans. The 32 intersectional strata ranged from two to 2,277 individuals. Four strata with fewer than 20 individuals were omitted from the I-MAIHDA procedure. The simple intersectional model had a VPC of 24.9%, indicating a substantial between-stratum variation in the lifetime prevalence of homelessness. Results of the additive main effects model indicated significant sociodemographic disparities in lifetime homelessness, most notably between Black and White participants (OR = 2.82, 95% C.I. = [2.46, 3.25]) and baby boomers and older generations (OR = 4.75, 95% C.I. = [4.14, 6.01]). The additive model had a VPC of 0% and a corresponding PCV of 100%, indicating that the included sociodemographic characteristics explained all between-stratum (intersectional identity) variance.

Conclusions and Implications:This study contributes to social work scholarship on homelessness by examining lifetime homelessness through an intersectional analytic framework. The large odds ratios associated with sociodemographic characteristics and a PCV of 100% indicated that individual sociodemographic characteristics, not intersectional identity, predict disparities in homelessness. These findings suggest that social work responses should affirmatively account for the structural inequities that drive long-standing racial disparities in homelessness and the increasingly relevant birth cohort effect. More broadly, this study demonstrates the value of intersectional quantitative methods in social work research for clarifying whether disparities arise from specific intersectional group identity or uneven risk across sociodemographic groups.