Date of Award
Summer 2025
Document Type
Open Access Thesis
Department
Psychology
First Advisor
Bret Kloos
Abstract
Interactions with the legal system are a crucial factor in maintaining homelessness. Difficulties in addressing legal issues are barriers to accessing services and finding employment and housing. Across the United States, the homeless court is an emerging legal program that has been developed to break this cycle, but they appear to face numerous capacity and coordination problems. Few empirical and systematic investigations have been conducted on homeless court programs. Further, previous studies of homeless courts focus their process and outcome measures on the priorities of the legal system. An investigation into the ecology, or environmental context, of homeless persons would improve our understanding of how such interventions transform the ecology, supporting improvements on the program model. As such, the proposed study aims (a) to investigate the effect of Columbia Homeless Court on its participants across income, living situation, health insurance and disabling conditions (b) test the feasibility of statistical learning methods for investigating the ecology of homeless court participants and non-participants. This quantitative study collected predictors and outcome data from the Homeless Management Information System on both participants of the homeless court program and others served by the Midlands Area Consortium for the Homeless. General linear models were used to compare outcomes between participants and non-participants with the homeless court. For their ease of interpretability and efficiency in identifying linear and non-linear relationships, tree-based statistical learning methods (classification trees, regression trees, and Bayesian additive regression/classification trees) were used to conduct variable selection and understand variable interactions to support theory development relating to the ecology of homeless court participants. Results indicated that (1) the assessment found null effects of the homeless court on the multiple tested outcomes and (2) the centrality of health insurance, age and income as selected variables was noted across all interpretable models. I also provide some recommendations to improve the feasibility of systematic homeless court evaluations and discuss the implications of the statistical learning models and their feasibility in understanding the ecology of homelessness.
Rights
© 2025, Dylan Wong
Recommended Citation
Wong, D.(2025). Investigating the Ecology of the Homeless Court: The Promise of Quasi-Experimental and Statistical Learning Methods. (Master's thesis). Retrieved from https://scholarcommons.sc.edu/etd/8382