Online Program

Reaching the Homeless in Australia: a longitudinal study

*Nicole Watson, University of Melbourne 
YiPing Tseng, University of Melbourne 

Keywords: homelessness, tracking, longitudinal, non-response analysis

As part of the Australian government’s commitment to halving overall homelessness by 2020, a research agenda was developed to expand the evidence base for understanding and preventing homelessness. A major component of this research is a new longitudinal study (Journeys Home: Longitudinal Study of Factors Affecting Housing Stability), which began in September 2011. Face-to-face interviews were conducted with 1676 individuals in the first wave and they will be followed at 6-montly intervals over a further 3 waves. This study will permit both the pathways into and out of homelessness to be examined as the sample will include people who are homeless as well as those vulnerable to homelessness.

Administrative data of income support recipients was used to select the sample. Since January 2010, people are flagged in the administrative data if they indicate they are without conventional accommodation (e.g., sleeping rough, squatting), living in temporary accommodation, or in medium- to long-term accommodation without a lease (e.g., boarding house or caravan park). Another group who have characteristics similar to those flagged but are not themselves flagged was identified via a modelling process. A clustered sample in 36 locations around Australia was selected from the flagged and non-flagged groups.

Tracking is a major exercise, even in the first wave as the contact details available from the administrative data may be considerably out of date for some people. The assistance of local organisations is sought, but privacy constraints limit the effectiveness of such avenues. Respondents are also asked for further contact details at the end of each interview to assist with the tracking efforts in later waves.

This paper will describe the sample design, tracking methods, and response rates to date. An analysis of non-response biases in the first wave will also be undertaken using the rich administrative data.