« Previous | Contents | Next »
Listen
Tracking Homelessness: A Feasibility Study
CHAPTER THREE: A REVIEW OF EXISTING LONGITUDINAL RESEARCH ON HOMELESSNESS IN THE UK AND US
3.1 This chapter reviews the (limited) existing longitudinal research on homelessness in the UK, supplementing this with selected US studies. We focus on those studies which attempt to fill at least one of the gaps identified in Chapter 2, namely prospective data on homeless people's progress over time and information on routes out of homelessness.
3.2 The studies reviewed fall into two main types: surveys (quantitative and qualitative); and administrative tracking data. While these two approaches to research can be combined in a single research study (as they were in O'Callaghan et al., 1996), they are more usually pursued independently and raise quite different methodological issues. For this reason they are discussed separately. The tables in Annexes 1 and 2 provide summaries of the key studies in the UK and US.
SURVEYS
3.3 Rossi (1990) identified three possible approaches to longitudinal research on homelessness:
(1) Full prospective - this is the most elaborate model. It would start with a sample of people at high risk of becoming homeless, contacting them periodically over a given time span, noting those who become homeless and, among them, those who remain homeless and those who leave homelessness. Rossi identified the major difficulty with this approach as identifying a group with a high enough risk of homelessness - the screening task would be formidable and much effort could be expended tracing people who do not experience homelessness.
(2) Semi-retrospective - this involves locating people who are already homeless and then following their progress over time, with their pre-homelessness experiences reconstructed by asking for retrospective accounts. While this approach is logistically easier than the full prospective model, Rossi highlights the potential unreliability of the retrospective data collected. There also remains the challenge of maintaining contact with a sample of homeless people, many of whom may be leading highly mobile and/or chaotic lives.
(3) Full-retrospective - Rossi describes this as the 'least attractive' but perhaps most feasible approach. Here retrospective accounts of entry and exit from homelessness are gathered from people who have been successfully re-housed. As well as the usual concerns about retrospective accounts, Rossi also points out that those people who have managed to move out of homelessness may not be representative of the entire homeless population, with the long-term homeless in particular likely to be under-represented.
3.4 We use this framework to structure our discussion of surveys.
Full prospective
3.5 Unsurprisingly, given the practical problems inherent in this type of research, full prospective research has seldom been attempted in either the US or UK. However, Stockley et al.'s (1993) tracking of 72 young people at risk of homelessness in the south of England could be said to fit into this category. These authors used agency and family informants, as well as more informal tracking through friendship networks, to locate and re-interview their sample one year after the initial interview. The concepts of 'maximum' and 'minimum' information were employed in this study, whereby 'maximum' denoted first-hand information received from the young people themselves (through interviews or questionnaire responses), and 'minimum' referred to second-hand data received from other informants (see also Smith & Gilford, 1991).
Semi-retrospective
3.6 Over the course of the 1990s the US has moved far ahead of the UK in relation to semi-retrospective research, with projects in a range of US cities following a sample of homeless people over time (Burt, 1999). 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. More recent studies tend to be both more ambitious and more successful in tracking homeless people.
3.7 A key early attempt at longitudinal follow-up was made by Sosin et al. (1990) who conducted a two-wave, six-month panel study of 451 homeless individuals recruited from homelessness services in Minneapolis (see also Piliavin et al., 1996). Their main research interest was patterns of exits and returns to homelessness. They used a variety of methods to locate sample members for the second interview, including agency and family/friend informants; change of address postcards; and signs and announcements in the relevant services. They also used modest cash incentives. They managed to re-interview 58 per cent of their sample, but do not comment on the relative success of the different search strategies employed.
3.8 In 1993, Cohen et al. published an extensive review of prospective homelessness research in the US, including detailed advice on tracking homeless people (discussed in Chapter 5.) They found that attrition rates varied widely in these studies, with some researchers managing to retain contact with only a third of their respondents, while others achieved follow-up rates of over 70 per cent. Cohen et al. attributed higher retention levels to more generous incentives, frequent contacts between interviews, and maintaining a presence in settings frequented by participants. In their own study of 163 homeless individuals, they managed to locate 86 per cent of participants at 4 months (re-interviewing 76%), and located 83 per cent at 12 months (re-interviewing 68%). The major variable distinguishing the locatability of respondents was service engagement, with those who accepted most help easiest to find. Even higher retention rates have been achieved in some recent studies, with Wong & Piliavin (2001) managing to re-interview 81 per cent of their sample of 564 homeless individuals up to a year after initial contact. Bassuk et al. (2001) successfully tracked 76 per cent of their sample of 220 homeless families over a period of two years. Unfortunately, neither of these sets of authors detail the tracking methods they used.
3.9 Survey research with a prospective dimension has less often been attempted in the UK, but several youth homelessness studies have included a follow-up element. Fitzpatrick (2000) used a similar approach to that of Stockley et al. (1993) in her detailed study of 25 young homeless people from Drumchapel in Glasgow. She managed to re-interview nine of these young people one year after the initial interview, and received a questionnaire response from another two. She obtained second-hand, 'minimum' information about a further 11 young people in her sample.
3.10 Craig et al.'s (1996) study of mental illness among young homeless people also included a panel element, but on a much larger scale. These researchers managed to successfully re-interview 107 young people out of their original sample of 161 after one year (a retention rate of 67%), and received second-hand information about the progress of a further 30 young people (19% of their sample). Their high level of tracking success seems attributable to the comprehensive and continuous tracking methodology used: "The follow up work began well before the appointed time of the second interview and was painstakingly thorough from the outset." (p.34) Craig et al. (1996) worked closely with a number of homelessness agencies, and these agencies were proactive in logging their contacts with young people in the sample and in encouraging them to arrange follow-up interviews with the researchers. Another 'vital' means of follow-up contact they used was networking through other homeless young people. The researchers found that asking social services and probation officers to forward letters to young people was less productive, as were attempts to trace the sample through GP registers. Sixteen young people took the initiative in keeping in touch with the researchers by letter or telephone.
3.11 At the other end of the age spectrum, Maureen Crane has carried out a series of sustained, ethnographic studies of the resettlement of older homeless people. Her most recent work has involved tracking the progress of 64 older people for a period of two years after they were re-housed into permanent accommodation (Crane, forthcoming). Crane has re-interviewed her sample every three or six months throughout this time, depending on whether they seemed to have 'settled' in their accommodation. Of the 17 people in her sample whose tenancies ended through eviction or abandonment, she managed to trace and re-interview ten. She traced them mainly through enquiries with agencies and other homeless people, but she searched the streets for some individuals. This was therefore a very intensive, individualised approach to tracking.
3.12 Most of the existing prospective research on homelessness in the UK consists of relatively small-scale, qualitative studies. By far the most substantial quantitative prospective survey has been the Department of the Environment funded study of homeless applicants to nine local authorities in England (O'Callaghan et al., 1996). A large sample of homeless applicants were interviewed as soon as possible after they made their application, with a subsequent interview taking place either when the applicants were permanently re-housed, or after their application was rejected or withdrawn. If the applicants were still awaiting decision at the end of the survey period (18 months), they were interviewed at that point. Out of 1,302 applicants eligible for second interview, 879 interviews were obtained (a 68% retention rate). A wide range of tracking methods were used in this study, but it was noted that local enquiries with former neighbours, contacts through friends or relatives, and checking local authority records for more recent addresses were the most successful. Change of address postcards and asking advice agencies to forward letters were found to be much less productive.
3.13 Important lessons on tracking homeless people can also be learned from prospective surveys with other 'hard-to-reach' groups. Harocopos et al. (forthcoming), for example, conducted a very interesting 18-month follow up study of 100 chaotic crack users in London. This sample was re-interviewed five times, with 72 respondents completing the final interview. The authors attribute this high retention rate to several factors, including their recruitment site (an agency which had a positive association for most users); strong rapport between researchers and respondents (built up through frequent interviews, continuity of researcher, sending birthday cards to sample etc.); and a 20 incentive. They mainly used contacts through family and friends to track down respondents, but also reported that treatment agencies and other services were highly co-operative because they were able to produce written consent from their respondents.
Full-retrospective
3.14 The limited review of US literature undertaken for this study did not cover full-retrospective research on homelessness, only studies in the UK were included. Vincent et al. (1995) surveyed the progress of men who had lived in Alvaston Resettlement Unit one year after it closed, and Tate et al. (2001) interviewed men resettled from the Great Eastern Hostel in Glasgow during its closure programme. Bennett (1990) carried out a very basic, but interesting, study of the re-housing experiences of 40 former residents of the Lochgelly hostel in Fife ten years after it closed. By then, 19 ex-residents were deceased and the whereabouts of six was unknown. Contact was made with 15 ex-residents, of whom 10 were interviewed. The researcher does not detail how the ex-residents were located, but notes that attempts to trace them carried on for nine months, and that local housing and social work agencies were 'excellent' sources of contact.
3.15 There are a small number of other examples of fully retrospective homeless research in the UK. For example, Randall & Brown (1996) conducted a study of ex-homeless people resettled in housing association accommodation through the Rough Sleepers Initiative (RSI), obtaining their addresses through the RSI Clearing House records. More recently, the 'routes out' research conducted for the HTF involved 27 in-depth, qualitative interviews with ex-homeless people who had sustained their accommodation for at least six months (Rosengard et al., 2002). It was designed as a study of 'successful' resettlement, with the sample selected to reflect a range of six potential routes out of homelessness hypothesised in the HTF pathways research (Anderson and Tulloch, 2001). The authors do not detail how they located the sample of ex-homeless people, though it is implied that they used local authority or homelessness agency contacts.
ADMINISTRATIVE TRACKING DATA
3.16 Interview surveys (both quantitative and qualitative) are often considered the 'ideal' means of collecting data on homelessness, because they generate first-hand information from respondents, and are based on questionnaires or topic guides tailored specifically to answer the research questions. However, Anderson & Tulloch (2000) argue that routine data collected by service providers also offers the potential to enhance longitudinal research approaches. Clearly, unique identifiers have to be allocated to each service user to enable longitudinal analysis of their movement through administrative systems, but local authorities and specialist homelessness projects are increasingly using such identifiers.
3.17 The largest administrative tracking study identified in this literature review was that of Wong et al. (1997) who monitored the exit and re-entry of almost 28,000 households to New York's family shelters over a two-year period. The database they used covered all of New York's family shelters, and recorded dates of admission, discharge, and subsequent re-admission, as well as housing destinations on discharge. The most substantial study in the UK to use administrative data to track homeless people was O'Callaghan et al. (1996) who tracked the progress of 2,474 homeless applicants for between 11 and 18 months. The authors reported that:
"The variety of administrative systems in place across the nine authorities, and the potential for a number of different housing department sections to be involved in each case, made the process of tracking applications an extremely difficult and time consuming exercise." (O'Callaghan et al ., 1996: 137)
3.18 Studies like Wong et al.'s (1997), and to a lesser extent O'Callaghan et al.'s (1996), should be relatively straightforward to conduct as they rely on manipulation of a single, unified database. A much greater challenge lies in linking records between separate homelessness databases and, even more ambitiously, linking records from the homelessness 'system' to that of other databases to track wider outcomes for (ex-) homeless people.
CONCLUSIONS FROM THESE STUDIES
Surveys
3.19 The existing surveys of homeless people both in the US and in the UK demonstrate their value in clarifying the processes involved in moving in and out of homelessness and, especially, in evaluating the impact of policy interventions. Burt (1999) argues that these studies have revealed the complexity of homelessness careers, and in particular have shown that many people experience single short spells of homelessness and manage to 'exit' on their own:
"These people may never draw much attention from service providers and planners because they do not draw heavily on service resources. Nevertheless, their experiences can help us understand the circumstances that allow people to leave homelessness and stay housed, and may also be important when planning prevention efforts." (Burt, 1999: 18)
3.20 These studies also confirm that the process of keeping in touch with homeless people is heavily resource intensive. Clearly the larger the sample size and the longer the tracking period, the more difficult it becomes to retain contact with a reasonable proportion of the sample. Retention rates are also likely to be affected by the type of homeless people being studied with, for example, older ex-homeless people resettled into specific accommodation (as in Crane, forthcoming) likely to be easier to track than young people leading chaotic lifestyles (although Harocopos et al. (forthcoming) indicates that high levels of retention can be achieved even with this group). The tracking methodology used also has an important impact, with the following points stressed in both the US and UK literature:
- it is crucial to obtain as many stable contact points as possible from each homeless person, including both agencies and friends/relatives;
- the same researcher should be used to track each homeless person throughout the study period as this helps to maximise rapport and commitment;
- flexibility and persistence are needed in tracking down all leads;
- telephone and face-to-face contact are usually more effective than written correspondence (although some studies have experienced success in sending out letters a week in advance of visits);
- frequent interviews (or low level contacts) are important in maintaining contact;
- it can be helpful to maintain a presence in the 'environmental niche' of the sample;
- incentives are required for some respondents; and
- a signed consent form is particularly helpful in encouraging agency informants to co-operate.
3.21 These points are covered in more detail in Chapter 5.
Administrative data
3.22 The existing studies using administrative data amply demonstrate both its strengths and weaknesses.
3.23 The strengths include:
- the very large samples that can be tracked;
- the relatively long timescales over which tracking can occur; and
- the opportunity it affords to avoid reliance on the recall of homeless respondents.
3.24 The weaknesses include:
- lack of researcher control over the scope of data collected, meaning that only a narrow set of research questions may be addressed;
- lack of rigour in agency record-keeping which can compromise data reliability; and
- the absence of evidence on outcomes for those who leave the 'system' and so drop out of the sample.
3.25 These issues are discussed in more detail in Chapter 4.
CONCLUSIONS
3.26 This chapter has reviewed the existing UK and US research which has a prospective dimension and/or offers data on routes out of homelessness. It found that these studies fell into two main categories: surveys and administrative tracking data. In relation to the former, a further division was suggested between full prospective, semi-retrospective and full-retrospective designs, based on Rossi (1991). The existing research demonstrates the strengths and weaknesses of both approaches, and provides a series of practical lessons for future research, especially in relation to keeping in touch with a cohort of homeless people. Chapter 4 uses this material to inform its discussion of the range of potential longitudinal research designs, and Chapter 5 details practical lessons on tracking homeless people.
« Previous | Contents | Next »