Community- and data-driven homelessness prevention and service delivery: optimizing for equity

Area of Science:

• Public Health
• Social Policy
• Data Science

Background:

• Federal policies mandate efficient and equitable local responses to homelessness.
• Limited homeless assistance resources necessitate data-driven decision-making tools.
• Prioritizing service allocation is challenging without empirical support.

Purpose of the Study:

• To test a community- and data-driven approach for homelessness prevention and resource allocation.
• To develop and evaluate prioritization rules for directing scarce homeless services.
• To assess the impact of these rules on reducing homelessness and reentry rates.

Main Methods:

• Utilized system-wide administrative records of homeless services and household reentry.
• Employed counterfactual machine learning to identify services preventing reentry.
• Integrated community input to aggregate predictions for subpopulations and create prioritization rules.
• Conducted simulations to compare rule-based reallocation with services-as-usual.

Main Results:

• Homelessness prevention showed benefits, with differential effects for various household types.
• Comorbid health conditions were best addressed by long-term supportive housing; families with children benefited from short-term rentals.
• Prioritization rules reduced simulated community-wide homelessness and mitigated reentry disparities for specific subgroups (e.g., female youth, unaccompanied youth).

Conclusions:

• Community- and data-driven prioritization rules enhance the equitable targeting of scarce homeless resources.
• Leveraging administrative data and machine learning supports local decision-making and service evaluation.
• This approach enables ongoing assessment of data- and equity-driven homeless services.