Academic journal article Australian Journal of Social Issues

How Many Australians Have Slept Rough?

Academic journal article Australian Journal of Social Issues

How Many Australians Have Slept Rough?

Article excerpt

Introduction

This paper investigates two issues: how many Australians have experienced homelessness during their lifetime, and how many people have slept rough. The two most commonly used sources of quantitative information about the homeless population are the Specialist Homeless Services (SHS) data collected by the Australian Institute of Health and Welfare (AIHW), and the quinquennial Census of Population and Housing undertaken by the Australian Bureau of Statistics (ABS). Both datasets provide important information on the characteristics of the homeless population. However, both datasets have limitations when it comes to examining the research questions that we are interested in. First, we explain those limitations, and then we outline an alternative methodological approach.

The SHS database collects information on all persons who request assistance from homelessness services over a 12-month period. This is an important database and it provides valuable information for policymakers and service providers. However, many homeless people do not use these agencies. One study found that only 40 per cent of homeless people had sought assistance from services while they were homeless (ABS 2011: 27).

The census counts those who seek assistance from services as well as those who do not, but census data has other limitations. One problem concerns the ability of census collectors to identify individuals who sleep rough. Census collectors are unlikely to find all of them because the census is held in winter when rough sleepers hide away to escape the cold, as well as for their own safety and security.

The second problem is that the census counts homeless people on a particular night. These counts are also referred to as 'point-in-time' counts or 'point prevalence' counts. Phelan and Link (1999) argue that point-prevalence counts do not provide an accurate picture of the homeless population. This is because those who have a long-term problem are more likely to be counted on census night than those who have a short-term problem. Phelan and Link (1999) refer to this as 'point-prevalence' bias:

As an illustration of 'point-prevalence' bias, imagine a survey conducted in a shelter on a given night in December. If residents come and go during the month, the number of residents on the night of the survey will be smaller than the number of residents over the month. If, in addition, length of stay varies, longer-term residents will be oversampled, and persistence will be overestimated (e.g., a person who stays all month is certain to be sampled, but a person who stays one night has a one in 31 chance of being sampled). Finally, if people with certain characteristics (e.g., mental illness) stay longer than others, the prevalence of those characteristics will be overestimated (Phelan and Link 1999: 1334).

This means that attempts to generalise about the social characteristics of the homeless population using census data are likely to be distorted. Research has established that the long-term homeless are more likely to have mental health issues and drug and alcohol problems than the short-term homeless (Culhane & Kuhn 1998; Johnson et al. 2013), and we know the long-term homeless will be over-represented in census counts. Conversely, the short-term homeless will be under-represented.

In seminal research carried out in 1990, Link and colleagues (1994; 1995) attempted to draw a sample of the homeless in the United States that adequately represented both the short-term and the long-term homeless. They undertook a national telephone survey (N=1507) of people who were housed and asked them if they had ever been homeless; our research builds upon their methodological insights. There are four advantages of the approach undertaken by Link and colleagues (1994; 1995). First, unlike the results of point-in-time counts, the short-term homeless will be adequately represented. Second, it is possible to establish how many people have a long-term and a short-term problem because their period of homelessness has ended, thus avoiding the problem of 'right censoring'--uncertainty about what happens to people after a particular time. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.