A Multivariate Analysis of Student Predictors of Homelessness

Background and Purpose: New York City homelessness prevention programs have incorporated risk models to be more efficient and effective in predicting homelessness. In this study we further advanced these risk models to identify students in NYC public schools at risk of future homelessness. Homelessness prevention programs have been recognized as a key intervention tool in addressing the social and familial circumstances associated with homelessness in New York City (NYC). The NYC Department of Social Services has integrated a predictive model into their intake process to better identify those most at risk of entering shelter through their homelessness prevention program. These efforts have led to a deeper understanding of the risk factors associated with homelessness and have been shown to be effective in targeting individuals and families most at risk. As we look upstream to prevent homelessness, NYC has identified risk factors based on foster care and educational administrative data to identify student level risk factors. The study aims to identify risk factors of imminent homelessness for NYC Department of Education (DOE) students and their families, and to evaluate the use of a predictive model for targeted interventions.

Methods: CIDI integrated administrative data from the NYC DOE, Administration for Children’s Services (ACS), Department of Youth and Community Development (DYCD) and Department of Homeless Service (DHS) to longitudinally examine student homelessness. The study population included students who were enrolled in NYC DOE schools any time between the 2011 and 2020 academic years. Student homelessness was defined as utilizing a DYCD or DHS homeless shelter and does not include street homelessness or other types of shelters. A logistic regression, clustered by students, was used to identify risk factors for student homelessness in the following school year among students who had never previously experienced homelessness. The model was then evaluated as a tool to predict imminent homelessness.

Results: This study identified several student-level risk factors that were strongly associated with the likelihood of experiencing homelessness in the following academic year. Black students were nearly 8 times as likely to experience homelessness compared to White students. The following factors approximately doubled a student’s risk of homelessness: attending school in the Bronx (compared to Staten Island), having participated in ACS prevention services, being overage for their grade, having ever been in a foster home, living doubled up with another family, transferring schools during the academic year, and being chronically absent.

Conclusions and implications: While the predictive model is robust statistically, the size of the group the model indicates is at high risk of homelessness is much larger than a prevention program could serve. Delivering prevention services to all students in the elevated risk group is not a viable intervention strategy. These findings imply that low-touch, wide-reaching school-based homelessness prevention may be a good first step to preventing homelessness, when paired with regular check-ins by a trusted and caring adult to identify the need for a more proactive approach.