| No.162/2003
Research Findings |
Development Department Research Programme |
Tracking Homelessness: A Feasibility Study
This document is also available in pdf format (103k)
Kevin Pickering, National Centre for Social Research, Suzanne Fitzpatrick, Department of Urban Studies, University of Glasgow, Kerstin Hinds, National Centre for Social Research, Sarah Tipping National Centre for Social Research and Peter Lynn, Institute for Social and Economic Research, University of Essex
Following work by the Homelessness Task Force which highlighted the lack of longitudinal research on homelessness, a feasibility study was commissioned by the Scottish Executive to consider ways in which longitudinal research might help address important policy questions and to identify practical methodologies for undertaking such research. The feasibility study considered practical issues of recruiting and retaining homeless people in longitudinal research as well as issues of Data Protection and ethics. |
Main Findings
The nature of this feasibility work means that the main findings predominantly take the form of a set of recommendations on practical approaches for studying homelessness experiences and outcomes over time. It is expected that these recommendations will be of particular interest for the work of the Homelessness Monitoring Group. (More details about these recommendations are shown in a table at the end of this Research Findings.)
- Consideration should be given to undertaking retrospective longitudinal research among ex-homeless people identified through the Scottish Household Survey (SHS). This would enable lessons to be learned about routes out of homelessness for both 'visible' and 'hidden' homeless groups.
- Questions should be added to the SHS, in order to enable the extent and nature of 'hidden' homelessness within Scottish households to be determined.
- At this stage it appears it would not be feasible to rely solely on administrative data to monitor homeless people's pathways. This is in part because administrative data would only allow very limited research questions to be answered. More fundamentally, the results would be biased towards those households that maintain contact with participating agencies, and no data would be available for those that drop out of the system. However, administrative data could be used both to obtain samples of homeless people and to assist in maintaining contact with people selected for any particular study.
- Relevant authorities should be encouraged to undertake prospective longitudinal studies of groups at high risk of homelessness - in particular, young people leaving care, people leaving prison and people discharged from mental health institutions. In such studies, exploration of experience of homelessness would ideally be just one element of a wider holistic study looking at the experiences of people after leaving these institutions.
- Longitudinal evaluations of specific programmes and projects should be developed to enable the long-term benefits to service users to be assessed. Some components of these evaluations should be standardised to allow comparison across programmes and projects.
- Encouragement should be given to a substantial longitudinal evaluation of the Glasgow Hostel Closure Programme, which is of a scale to allow robust quantitative comparisons of outcomes across different types of new provision.
- Local authorities should be asked to report in their homelessness strategies how they will use and encourage the use of longitudinal methods to assess the impacts of projects and policies they have pursued in tackling homelessness.
- The lessons learned in previous longitudinal studies in the UK and the US (where more such work has been undertaken) should be considered in developing any future longitudinal research on homelessness, with particular attention paid to the recommended techniques for 'tracking' samples of homeless people over time.
- In order to meet the requirements of the Data Protection Act 1998, appropriate 'permission to locate' and/or consent forms should be completed for any prospective longitudinal research study.
Methods
A review of relevant literature from the UK and the US was undertaken to assess what research methods had been used in previous longitudinal research into homelessness and with what degree of success. The potential for longitudinal research to address outstanding policy questions in this area was also considered.
In parallel, information currently recorded in Scotland on homelessness was assessed for its potential to facilitate longitudinal research. This largely comprised administrative data from local authorities and voluntary organisations working in the homelessness field, as well as data from large Scottish population surveys.
Based on the literature and data reviews, the feasibility of a range of potential strategies for longitudinal research with homeless people was assessed. Telephone interviews with people holding data and with people who had carried out relevant research were undertaken as part of this process.
Data protection and ethical issues were also considered for all of the potential designs.
Findings from the Policy and Literature Reviews
The review of policy needs identified two main potential contributions of longitudinal research on homelessness. First, such research could enable a much more precise identification of the circumstances which lead households to move in and out of homelessness, enhancing the evidence base for both prevention and resettlement work. Second, longitudinal data would allow the long-term impacts of particular homelessness projects and programmes to be assessed.
The literature review aimed to include all homelessness research in the UK with a longitudinal dimension, and a selection of relevant US studies was also reviewed. Longitudinal studies in both the UK and US fell into two main categories: surveys and studies relying on administrative tracking data. Three possible approaches to longitudinal research on homelessness using surveys were identified (Rossi, 1990):
- full prospective - following-up a sample of people at high risk of becoming homeless, and noting those who become homeless and, among them, those who remain homeless and those who leave homelessness over time.
- semi-retrospective - obtaining a sample of people who are already homeless and then following their progress over time, with their pre-homelessness experiences reconstructed by asking for retrospective accounts.
- full-retrospective - collecting retrospective accounts of entry into and exit from homelessness from ex-homeless people.
The full prospective approach is seen as the 'gold standard' design, but would be expensive and difficult to undertake (if not impossible in many cases). For most studies, the semi- and full-retrospective approaches would be more feasible, although both would be subject to recall error associated with retrospectively collected data.
No full prospective studies of homelessness have been undertaken in Scotland (and only one in the UK) and the number of semi- and full-retrospective studies is also limited. Most research in the UK has relied on retrospective accounts from people currently living in 'visible' homeless situations. Because of this, much more is known about routes into homelessness than routes out, and very little is known about the experiences of 'hidden' homeless people.
In the US, full prospective research has also seldom been attempted. However, the US is far ahead of the UK in relation to semi-retrospective research. Typically, these US studies comprise large-scale, quantitative surveys which repeatedly interview a cohort of individuals or families sampled from homelessness services. Tracking time scales range from six months to two years, and there are usually two or three waves of interviews.
Some large-scale administrative tracking studies have been undertaken in the US (there has been only one in the UK - O'Callaghan et al., 1996). These studies have identified the strengths of using administrative data to 'track' people - namely the large sample sizes and the potentially long follow-up periods. However, they have also demonstrated the weaknesses - mainly the limited number of research questions that can be studied and the bias caused by 'lost' cases.
Review of the Data Sources
The review of the data sources currently available in Scotland considered both surveys and administrative data and identified a number of key potential sources of longitudinal data on homelessness.
Of the large-scale surveys currently undertaken in Scotland, only the Scottish Household Survey (SHS) could feasibly contribute to a longitudinal study of homelessness. This is mainly because it contains questions about experiences of homelessness - each year about 550 SHS respondents report that they have had some experience of homelessness in their lifetime.
Local authorities collect a wealth of information about households that apply for housing as homeless for the HL1 returns. This information is collated by the Scottish Executive and constitutes the most comprehensive single dataset on the homeless population in Scotland. At present, certain groups of homeless people, in particular single homeless people, are likely to be under-represented in this database (Fitzpatrick et al., 2000). However, because of recent and proposed changes to the homelessness legislation in Scotland, the coverage of the HL1 returns dataset is likely to improve in the future. HL1 returns could prove a useful sampling frame for identifying people homeless at a particular point in time who might then be tracked over time.
Other large administrative datasets are maintained by the Glasgow Homelessness Network (GHN) to monitor the Rough Sleepers Initiative (RSI) and by Shelter Scotland. They contain information about people that contact any RSI project and Shelter respectively. Both could be used to obtain samples of homeless people, assuming appropriate consent could be obtained. This would be useful to supplement any sample(s) selected from the HL1 returns, so that both single homeless people and people sleeping rough would be better represented.
For the reasons outlined above, it does not appear feasible to rely solely on administrative data to monitor homeless people's pathways at this time. However, with the appropriate consents, such data could be very useful in helping to find and re-contact respondents who had agreed to take part in a longitudinal study.
Research Approach | Type of research | Types of homeless people | Positive or negative outcomes tracked | Type of data generated | Concerns |
Following up people who report experience of homelessness in the SHS | Full retrospective | Ex-homeless. 'Hidden' homeless (especially if relevant questions added). All household types - including families. | Positive | Qualitative (and possibly quantitative) follow-up study | Permission to re-contact rate relatively low |
Adding further questions on homelessness to the SHS | Full retrospective | Ex-homeless. 'Hidden' homelessness (especially if relevant questions added). All household types - including families. | Positive | Quantitative | None |
Sampling from administrative data | Semi- retrospective | All 'visible' groups of homeless households - including families. | Both | Quantitative | Data protection issues |
Monitoring pathways using administrative data | Semi- retrospective/ prospective | All 'visible' groups of homeless households - including families. | Depends on breadth of data sources used | Quantitative | Bias from 'lost' people. Data protection issues. Difficulties in linking datasets |
Studying at risk groups e.g. those leaving care or prison | Full prospective | Depends which risk categories targeted - probably mostly single homeless. | Both | General quantitative survey of selected groups. Follow-up qualitative survey if experience of homelessness. | Obtaining the sample. Would need policy interest in general survey of selected population. |
Evaluation of projects | Semi- retrospective | 'Visible' groups. Probably mainly single homeless. | Both. | Quantitative & qualitative (probably mainly the latter). | Need to ensure common elements in evaluations to allow for comparisons |
Research Designs
From the results of the literature and data reviews, the research team identified a number of potential designs for studying homelessness. Those considered most valuable were developed in some detail. The designs that were recommended or rejected were outlined on the first page of this Research Findings. The table on page 3 summarises the full range of potential designs considered in detail and shows for each research approach: the type of longitudinal research that could be undertaken; the types of homeless people that could be studied; whether the design best covers people moving out of homelessness (positive outcomes) or remaining in/moving into homelessness (negative outcomes); the likely type of data generated; and any concerns with the design.
If you wish further copies of this Research Findings or have any enquiries about social research, please contact us at: Scottish Executive Social Research
1F (Dockside)
Victoria Quay
EDINBURGH
EH6 6QQ Tel: 0131 244 7565
Fax: 0131 244 7573
Email: socialresearch@scotland.gsi.gov.uk
Web site: www.scotland.gov.uk/socialresearch The report "Tracking Homelessness: a Feasibility Study", which is summarised in this Research Findings is available on the Social Research Website http://www.scotland.gov.uk/socialresearch This document (and other Research Findings and Reports) and information about social research in the Scottish Executive may be viewed on the Internet at: http://www.scotland.gov.uk/socialresearch The site carries up-to-date information about social and policy research commissioned and published on behalf of the Scottish Executive. Subjects covered include transport, housing, social inclusion, rural affairs, children and young people, education, social work, community care, local government, civil justice, crime and criminal justice, regeneration, planning and womens issues. The site also allows access to information about the Scottish Household Survey. |