Using Linked Data to Examine the Housing Trajectories and Social, Educational, and Health Outcomes of Families and Children Experiencing Homelessness

Background & Purpose 

This study examines family homelessness in Charlotte-Mecklenburg County, North Carolina from 2014-2016 using linked data from local county sources and the US Census data infrastructure.  The purpose of the study is 1) to examine the longitudinal housing trajectories of families experiencing homelessness, 2) to compare these patterns with similar families who have not been homeless, and 3) to identify risk and resilience factors for families and children with respect to their interactions with housing agencies, the child welfare system, the county sheriff’s office, and child academic and behavioral outcomes collected by county schools.  It is hypothesized that families experiencing homelessness will be at increased risk for negative social, behavioral, and health outcomes.

This study is unique as it links county-held data with the data infrastructure of the US Census, primarily the American Community Survey (ACS), along with data on housing and assistance programs.  Linking these data allows for the development of a comparison group of families who are similar with respect to demographic, socioeconomic, and cultural factors to those who did experience homelessness but maintained housing stability, and allows for the tracking of housing trajectories after an episode of homelessness – data often elusive to local and county agencies and researchers alike.  

Methods

The sample includes 1,820 families – 910 who experienced homelessness, 910 who have not.  Because the treatment condition (i.e. homelessness) is determined through non-random assignment, there is a risk that such an approach would suffer from selection bias. To address this issue, propensity scoring (PSM) is used to allow for causal inference in cases like these where random assignment is not possible.  PSM has also been previously used in homelessness research to compare samples in studies without randomization.

Three sets of analyses are conducted to answer study research questions.  First, longitudinal descriptive analyses are used to examine the housing trajectories of families experiencing homelessness.  Next, propensity score matching is used to compare the effect of homelessness on subsequent housing trajectories.  Finally, multilevel regression analyses are used to develop predictive models that identify factors associated with 1) increased risk for subsequent homelessness, 2) risk for child welfare involvement, 3) risk for criminal justice involvement, and 4) resilience factors associated with improved academic outcomes in schools for children in families experiencing homelessness.

Results

Initial findings indicate that the majority of families experiencing homelessness in Charlotte are not unsheltered but access services through transitional housing and emergency shelters. Children in families experiencing homelessness are at greater risk for reading below proficiency, chronic absenteeism, and school suspension.  Roughly a third of children in transitional housing or emergency shelter are not accessing McKinney-Vento services.  Additional findings highlight the complex interaction of race, ethnicity, and gender with risk factors including exposure to violence, childhood trauma, and poverty. Final results for this study are forthcoming. 

Conclusions & Implications

Study findings highlight how traditional forms of housing support will need to work in collaboration with other agencies and sectors serving families and children to support wellbeing for this population.