Using linked data for rigorous outcomes evaluation of social housing programs

Abstract

In Australia there is very limited supply of social housing with long waitlists. In 2016 the New South Wales government launched major reforms aiming to address the availability and quality of social housing stock, to better assist social housing clients in gaining financial independence and improve the experience of clients. We worked in a consortium to carry out an evaluation of three programs delivered under these reforms: A fixed term subsidy to support households to rent in the private rental market (a diversion from social housing) Case management supports to help housing clients find or increase their employment A scholarship fund to help disadvantaged students with education-related expenses The evaluation used rigorous mixed-methods approaches and was ground-breaking in creating a large, linked data asset to track outcomes of clients over time. The linked dataset spanned the population of housing and homelessness clients across housing, homelessness, health, education, justice, welfare and child protection government datasets. We used this dataset to carry out outcomes evaluations for each program using quasi-experimental designs: propensity score matching, stepped wedge models and regression discontinuity respectively. The linked data was critical to the evaluation allowing us to: Better understand client needs based on past service use Use quasi-experimental designs to robustly measure program impacts Measure outcomes across a range of domains (e.g. health and justice). In this talk we will cover the use of the linked dataset including the quasi-experimental designs and outcome indicators, as well as the findings of the evaluation.