Factors contributing to becoming housed for women who have experienced homelessness.
Article Type: Report
Subject: Homeless women (Care and treatment)
Homeless women (Social aspects)
Homeless shelters (Research)
Authors: Nemiroff, Rebecca
Aubry, Tim
Klodawsky, Fran
Pub Date: 12/22/2010
Publication: Name: Canadian Journal of Urban Research Publisher: Institute of Urban Studies Audience: Academic Format: Magazine/Journal Subject: Social sciences Copyright: COPYRIGHT 2010 Institute of Urban Studies ISSN: 1188-3774
Issue: Date: Winter, 2010 Source Volume: 19 Source Issue: 2
Topic: Event Code: 290 Public affairs; 310 Science & research
Product: SIC Code: 8361 Residential care
Geographic: Geographic Scope: Canada Geographic Code: 1CANA Canada
Accession Number: 254482910
Full Text: Abstract

This longitudinal study examined physical integration, defined as becoming rehoused, of women who were homeless at the study's outset. Participants (N = 101) were recruited at homeless shelters in Ottawa and participated in two in-person interviews, approximately two years apart. A predictive model identifying factors associated with becoming rehoused and achieving housing stability was developed from previous empirical research and tested. Being accompanied by dependent children and having access to subsidized housing predicted being re-housed at follow-up. This research represents the first longitudinal study examining exits from homelessness in a sample of Canadian women. The findings suggest that providing financial resources is essential to helping women who have experienced homelessness to become physically integrated in their communities. In addition, it is suggested women who are unaccompanied by children would benefit from more intensive services.

Keywords: homelessness, housing, community integration, women

Resume

Cette etude longitudinale a examine l'integration physique dans la communaute, un concept qui fait reference au relogement des femmes qui etaient sans-abri au moment de commencer l'etude. Les participants (N= 101) ont ete recrutes dans les refuges pour sans-abri a Ottawa et ont participe a deux entrevues a deux, a environ deux ans d'intervalle. Un modele predictif identifiant des facteurs associes au processus de relogement et a la realisation de la stabilite du logement a ete elabore a partir de recherches empiriques anterieures; le modele a aussi ete teste. Etre accompagne par des enfants a charge et avoir acces a un logement subventionne prevoyait le relogement au moment du suivi. Cette recherche constitue la premiere etude longitudinale examinant les issues de l'itinerance, au sein d'un un echantillon de femmes canadiennes. Les resultats suggerent que de fournir des ressources financieres est indispensable pour permettre aux femmes qui ont vecu l'itinerance a devenir physiquement integrees dans leurs communautes. En outre, il apparait que des femmes qui ne sont pas accompagnees d'enfants pourraient beneficier de services plus intensifs.

Mots cles: itinerance, logement, integration communautaire, femmes

Introduction

Homelessness is a growing problem in Canada, and one that is garnering increasing attention in both the research literature and public consciousness (Gaetz, 2010; Hulchanski, Campinski, Chau, Hwang and Paradis, 2009). Recently, the United Nations Special Rapporteur on adequate housing described the situation in Canada as a "national crisis" and noted that women are among those who are particularly vulnerable to difficulties associated with inadequate housing (United Nations, 2007).

People who are homeless may be disaffiliated from the mainstream of society and often face barriers to becoming reintegrated in their communities. Clapham (2003) points out the importance of housing in allowing the normal activities of living, for example work and family life, to occur. Breese and Feltey (1996) argue that becoming homeless means the loss not only of housing, but also of the role of housed citizen and fully-functioning member of society.

This study examined predictors of becoming housed for women who have experienced homelessness from the perspective of community integration. Research to date investigating successful exits from homelessness leading to physical integration in the community have been conducted exclusively in the United States, with much of it conducted in the 1990s. The objective of the present study was to examine which factors contributed to women leaving homelessness in a Canadian context in the 2000s.

Homeless Women and Families

Homeless women, whether alone or with children, face a diverse set of challenges, including physical illness, low levels of education, unemployment, victimization, (Buckner, Bassuk and Zima, 1993; Fisher, Hovell, Hofstetter and Hough, 1995) and, frequently, histories of family disruption and violence in childhood (Farrell, Aubry, Klodawsky, Jewett and Petty, 2000; Shinn, Knickman and Weitzman 1991; Shinn et al., 1998). Women who are homeless report higher levels of psychological distress and mental health problems than homeless men (Roll, Toro and Ortola, 1999). Women who are unaccompanied by children are more likely than women with dependent children, but less likely than single men, to report substance abuse difficulties (Farrell et al.; Roll et al.; Zlomick, Robertson and Lahiff, 1999). Homeless women report higher levels of social support than do homeless men (Farrell et al.), but may also have less work experience and fewer work skills (Roll et al.). Women with dependent children appear to experience the fastest and most stable exits from homelessness. In contrast, women who are unaccompanied by children, while more likely than men to exit homelessness, are also more likely to experience repeated episodes of homelessness (Piliavin, Wright, Mare and Westerfield, 1996; Zlomick, Robertson and Lahiff).

Community Integration

Social Role Valorization theory (SRV) focuses on the importance of social roles to community integration. According to SRV, those who hold valued social roles are more likely than those who do not to get "the good things in life," including access to material goods, needed services, and decent housing, as well as respect, acceptance, positive relationships, and integration into valued activities and social functions. Those who hold devalued roles are likely to receive the opposite: poorer quality food, housing, clothing, education, and health care, work others do not want, violence and brutality, scapegoating, rejection, separation, segregation, and exclusion (Thomas and Wolfensberger 1999).

Although SRV theory was developed to conceptualize community integration of people with developmental disabilities, it can be easily applied to people who are homeless. Homeless individuals are placed in roles devalued by society; they have fewer opportunities and less access to material goods than most people, and are frequently victimized. Being cast in devalued roles limits homeless people's access to valued roles, such as that of a worker, parent, or competent person. Thus, they often remain in devalued roles, which, in turn, reinforces the low value society ascribes them. Individuals may begin to identify with these negative roles, thus becoming further entrenched in the culture of homelessness and devaluing themselves (Farrington and Robinson 1999, Grigsby Baumann, Gregorich and Roberts-Gray 1990, Snow and Anderson 1987).

Several facets of community integration have been identified in the literature. Aubry and Myner (1996) identified three facets of community integration. Social integration includes normative interactions with community members, and the size, diversity, and support provided by individuals' social networks. Psychological integration is a sense of belonging in the community. Aubry and Myner (1996) defined physical integration as participation in the community outside the home. For people who are homeless, physical integration may be defined as being stably housed: to become present in the community is not to become fully integrated, but it is difficult to become re-integrated until a physical presence has been established among the housed population. Storey (1989, cited in Flynn and Aubry, 1999) argued that for individuals with disabilities, physical integration is the "necessary first step for other forms of integration" (p. 276).

Little previous research has examined becoming re-housed following homelessness from the perspective of community integration. No empirical studies were found examining the community integration of a non-clinical sample of people who have experienced homelessness; however, one research group has investigated the community integration of people with severe mental illness who have experienced homelessness. These authors found that having more choice in housing and living in scattered, independent housing, rather than institutional settings, were associated with higher levels of both social and psychological integration (Gulcur, Tsemberis, Stefancic and Greenwood, 2007). Participants reported that being housed was a normalizing experience; they described feeling "normal" and "part of society" as a result of being housed (Yanos, Barrow and Tsemberis 2004).

Becoming housed may be the first step toward becoming integrated in the community. Guest and Stamm (1993) found that finding housing and work were the first priorities of individuals moving to a new city. It is also important that housing be stable in order for individuals to become truly integrated in the community. Moving from homelessness to low quality housing and back again may present an important obstacle to other forms of community integration. As individuals struggle with poverty and unstable, sub-standard living conditions, they continue to occupy devalued social roles, excluding them from valued social roles and becoming full participating community members. The present study tested a model of physical integration in the community for a sample of women who have experienced homelessness.

Model of Physical Integration

The model of physical integration guiding our study is based on the work of Piliavin and colleagues (1993, 1996) and Zlomick and colleagues (1999). As presented in Figure 1, the model specifies several predictors to two aspects of physical integration, Housing Status (i.e., housed or not housed) and Housing Stability (i.e., length of time housed). Our review of the literature on homelessness among women found support for the predicted variables of leaving homelessness presented in the model.

[FIGURE 1 OMITTED]

Family Status

Family status was conceptualized based on whether participants were accompanied by dependent children at any time over the study period. Past research has demonstrated that women with dependent children tend to have shorter episodes of homelessness than women who are unaccompanied by children, (Burr and Cohen 1989, Wong and Piliavin 1997, Zlomick, Robertson and Lahiff 1999) and are more likely to obtain stable housing (Wong, Piliavin and Wright 1998, Zlomick, Robertson and Lahiff 1999).

Disaffiliation

Disaffiliation was conceptualized as having low levels of perceived social support. A number of indicators of disaffiliation have been studied in relation to homelessness, including disruptive childhood experiences, parental and marital status and criminal history (e.g. Piliavin et al., 1996, Zlotnick, Robertson and Lahiff 1999). However, only perceived social support has been associated with exiting homelessness (Cohen Ramirez, Teresi, Gallagher and Sokolovsky 1997) and achieving housing stability (Anucha 2003, Bassuk et al., 1997).

Acculturation to Homelessness

Acculturation to homelessness was operationally defined by the amount of time spent homeless in a participant's lifetime, with longer lifetime histories of homelessness representing higher levels of acculturation. Piliavin and colleagues (1993) found an association between comfort with street life and total amount of time spent homeless. Zlotnick and colleagues (1999) used the total amount of time spent homeless over a lifetime as a measure of acculturation. Shorter lifetime histories of homelessness have been associated with an increased chance of exiting homelessness and obtaining stable housing (Piliavin et al., 1996, Zlomick, Robertson and Lahiff 1999).

Personal Disability

Personal disability was examined in terms of mental health difficulties, specifically symptoms of anxiety and depression and consequent impairment in functioning. Symptoms of specific mental illness were not examined, as severe mental health symptoms were expected to cause significant difficulties in functioning, which, in turn, would be captured by a measure of general impairment. Mental health difficulties have been associated with a decreased chance of exiting homelessness for women unaccompanied by children, and a return to homelessness for women with dependent children (Wong and Piliavin 1997).

Human Capital

More time spent in the work force has been associated with a decreased chance of returning to homelessness after becoming housed (Piliavin et al., 1996). Caton and colleagues (2005) found that a having a history of current or recent employment predicted shorter episodes of homelessness. In a qualitative study, Anucha (2003) found that participants cited unemployment and underemployment as important challenges to housing stability.

Housing Support

Housing support, such as receiving subsidized housing, has been found to be an important predictor of becoming and remaining housed in a number of studies (e.g. Zlomick, Robertson and Lahiff 1999, Wong and Piliavin 1997, Shinn et al., 1998). Receipt of government assistance such as welfare has been found to facilitate housing stability in American studies (e.g. Zolnick et al., Shinn et al.), but as the vast majority of participants in the current research were receiving benefits at the time of the initial interview, the role of government assistance was not examined.

The current study expands on previous research on homelessness and becoming re-housed following homelessness, and examines this experience from the perspective of community integration. Few studies have been conducted in the Canadian context, and most existing studies were conducted in the 1990's. In addition, by using longitudinal data, this research contributes to a small, but growing, body of work concerned with following individuals who have experienced homelessness over time.

Method

Participants and Procedure

"This research was conducted as part of a larger, longitudinal study on homelessness in Ottawa (Aubry, Klodawsky and Hay 2003, Aubry, Klodawsky, Nemiroff, Birnie and Bonetta 2007). "The methodology used in the study was approved by the Research Ethics Board at the University of Ottawa. Data collection took place in two phases; the fall and winter of 2002-2003 and the fall and winter of 20042005. At Time 1, approximately equal numbers of participants were recruited in each of five subgroups: adult men who where unaccompanied by children, adult women who were unaccompanied by children, adults with dependent children, male youth, and female youth. The sample for this paper consists of all adult women aged 20 or older who participated in both phases of the larger study (N = 101). All participants were homeless and resided at emergency shelters at Time 1. Criteria for homelessness, based on Toro and colleagues (1999), were that participants did not have their own house, apartment, or room and were either a) living on the streets or in shelters or b) staying temporarily in the home of friends or family and did not pay rent regularly.

At Time 1, the sample consisted of 85 women who were unaccompanied by children and 69 women with dependent children. A quota sampling strategy was used in order to select a sample that was representative of the population of shelter users in Ottawa. Stratification was based on population data provided by participating shelters. For women unaccompanied by children, the sample was stratified in terms of length of shelter stay in the following proportions: under 14 days (36%); 14-26 days (18%); 27-61 days (21%); 62 or more days (21%); and unknown (4%). The sample was also stratified in terms of citizenship (Canadian or other), and approximately 85% of women unaccompanied by children were Canadian citizens. There was no stratification based on length of shelter stay for women with dependent children, and 70% of the women with dependent children were Canadian citizens. The number of individuals recruited at each shelter was also proportionate to the number of potential participants residing at each shelter. Shelters sampled included city-run family shelters, a general-purpose women's shelter, shelters for women fleeing domestic violence, a shelter for newcomers to Canada, and a shelter for Aboriginal women. Additional eligible participants were recruited from other services, including drop-in centres and meal programs. Program staff aided in recruiting participants. Participants received honoraria of $10.00 for participation in the first interview and $20.00 for the second interview.

In order to locate participants for follow-up interviews participants were asked at Time 1 to provide consent for researchers to contact individuals and/or agencies including friends, family members, service providers, hospitals, and shelters, as well as to provide their personal cell phone numbers and e-mail addresses when available. Participants were also asked to provide consent for the researchers to contact the municipal Employment and Financial Assistance branch, which distributes social benefits, to search for addresses and phone numbers of participants.

The majority of participants signed both forms, and provided at least one contact person. Brief follow-up was made by phone approximately one year following the initial interview, to maintain contact and update contact information (Aubry, Klodawsky, Hay, Nemiroff and Hyman 2004). Participants were re-contacted approximately one year later to schedule follow-up interviews. Retention rates were 71% for women with dependent children and 65% for women unaccompanied by children. The final follow-up sample consisted of approximately equal numbers of women with dependent children (n=49) and women unaccompanied by children (n=52).

Measures

Predictor Variables

Family Status. Participants were asked whether they had any children, the children's ages, and how many of their children normally lived with them at both Time 1 and Time 2 interviews. Women with dependent children were those who lived with dependent children under age 18 at any time during the study period. All other participants were unaccompanied by children. Participants who resided in family shelters at Time 1 were considered women with dependent children since they had at least one dependent child under 16 years old living with them. Family status was confirmed at Time 2 using Toro and colleagues' (1997) Housing, Income, and Service Timeline (HIST), which consists of a housing history table on which the interviewer records a detailed history of the participant's living situation, employment, and income.

Employment History. Employment history was operationally defined as the percentage of time the participant had spent in the work force since age 16. This was assessed via an interview question at Time 1: "Since you were 16 years old, approximately how many years have you spent working for pay?" A proportion of adult years spent working was then calculated. First, the number of years since age 16 was calculated by subtracting 16 from the participant's current age at Time 1. The number of years worked was then divided by the number of years since age 16 in order to derive a proportion.

Mental Health Functioning. Mental health functioning was assessed at Time 1 using the mental component summary measure (MCS) of the SF-36, Version 2 (Ware, Kosinski and Gandek 2002). The SF-36 is a screening instrument that provides a self-report measure of physical and mental health relative to a general population that can be matched by age and sex. Ware and colleagues report internal consistency and reliability for group comparisons above .80 for all sub-scales of the SF-36. For the current sample, internal consistency for the MCS was high ([alpha] = .92).

Perceived Social Support. Perceived social support was measured at Time 1 using a brief, 5-item version of the S Scale of the Social Support Questionnaire (SSQ) (Sarason, Levine, Basham and Sarason, 1983). On the measure, participants list individuals who provide them with various types of support, then rate their satisfaction with each type of support on a 6-point scale that ranges from 1 ("very dissatisfied") to 6 ("very satisfied"). A sample item from the SSQ is, "Whom can you count on to console you when you are very upset?" q-he overall S score is obtained by dividing the sum of scores for all items by 5. q-he possible range of scores for the 5-item SSQ s Scale is from 1-6, with high scores indicating higher levels of satisfaction. Sarason and colleagues report very high internal consistency of the full-scale, 27-item SSQ (a = 0.94 for S). Test-retest reliability for the full scale over a period of one month is also strong (r = .83 for S). For the current sample, internal consistency of the 5-item version was also high, [alpha] = 0.87.

Lifetime History of Homelessness. Number of months spent homeless was measured using the HIST, described above (Toro et al., 1997). At Time 1, participants were asked their housing history for the past three years. Participants were then asked if they had any additional experiences of homelessness that had not already been recorded, and if so, how long they had been homeless. These questions were repeated until participants indicated that they had experienced no additional episodes of homelessness. Further episodes of homelessness were recorded at Time 2. If additional episodes of homelessness occurred between the Time 1 and Time 2 interviews, these were included in the calculation of the final amount of time homeless for each participant. However, the most recent homeless episode before becoming housed was excluded in order to avoid overlap with the housing stability outcome variable. Toro and colleagues report a test-retest correlation of the time homeless variable of the HIST of .73, based on a lifetime version of the HIST in a study of 31 homeless adults.

Housing Support. Housing support was a dichotomous variable based on whether participants accessed subsidized housing at any time over the follow-up period. Housing support was assessed via the HIST (Toro et al., 1997) at Time 2. Participants were asked, "What type of housing was it?" for each address they reported. If participants reported living in subsidized or social housing at any address listed at Time 2, they were considered to have accessed subsidized housing. In addition, addresses of housing in which participants lived were compared to a current listing provided by the municipal government of social housing addresses in which rent was subsidized. Finally, participants who responded "yes" to either of the following questions about their current housing at Time 2 were considered to have accessed subsidized housing:

1. Do you know whether the place you are living in now is social housing?

2. Is the rent for your current housing subsidized?

Physical Integration Outcome Variables

Two aspects of physical integration, housing status and housing stability, were examined in order to understand participants' experiences with becoming rehoused.

Housing Status. A dichotomous variable was created to classify participants as "housed" or "not housed." Participants were considered housed if they had been in housing for 90 days or more at follow-up. Ninety days was chosen as it reflects a moderate amount of housing stability, and is used as a measure of housing stability by the Government of Ontario. Housing was defined as owning or renting one's own house, room, or apartment, including subsidized or social housing, or living on a permanent basis with other people such a spouse, partner, or family members. Participants were considered housed if they had been living in transitional, supportive, or supported housing for at least 90 days at follow-up. Housing status was assessed at Time 2 using the HIST (Toro et al., 1997) described earlier.

Housing Stability. Housing stability was operationally defined as the amount of time that the participant had been housed at follow-up. The HIST (Toro et al., 1997) was used to ascertain the number of consecutive days participants were housed at Time 2. Since the length of the follow-up period varied between participants, housing stability was calculated for each participant by dividing the number of days consecutively housed at Time 2 by the number of days between interviews.

Results

Sample Characteristics

A total of 154 women were recruited at Time 1. The current sample consists of the 101 participants who participated in the follow-up interview. These were 52 women unaccompanied by children and 49 women with dependent children. At Time 1, the average age of participants was 35.6 years (SD = 10.5). The majority (76%) reported being non-partnered (i.e., single, separated, divorced, or widowed), while 24% reported being partnered (i.e., married, living with a romantic partner, or being in a common-law relationship). Participants reported having, on average, 1.6 children under age 18 (SD = 1.46). Women with dependent children had more minor children on average than women unaccompanied by children, t (98) = -6.07, p < .001, two-tailed. However, a substantial minority (44%) of women unaccompanied by children did report having at least one child under age 18.

A significant minority of participants (37%) reported that they had less than a high school education. Of the remainder, 43% reported a high school diploma as their highest educational attainment, 11% reported having a college diploma or trade or technical certificate, and 8% reported having a university degree. Thirty-eight percent (n = 59) reported that they were not born in Canada. Of these, 41% reported being Canadian citizens, 31% reported being refugees, and 21% reported being landed immigrants.

Table 1 shows means and standard deviations or percentages for the sample on each of the variables in the physical integration model. Overall, the women in this study reported low levels of mental health functioning at Time 1 (M = 35.60, SD = 12.54) compared to the normative sample of respondents from the 1998 survey of the U.S. general population (M = 50, SD = 10) (Aubry, Klodawsky, and Hay 2003). No Canadian comparative data were available. At Time 2, participants reported being homeless for, on average, approximately 19 months over their lifetimes; however, there was a large degree of variability in this (range = <1 month to 152 months). Many participants (20%) reported additional episodes of homelessness during the follow-up period.

A majority of participants (68%) reported being on the waiting list for social housing at Time 1. Among this group on the waiting list, 80% reported accessing subsidized housing during the study period at Time 2. Participants who had been housed for at least 90 days (n = 89) at Time 2 reported spending, on average, 36% of their income on housing. Of those who were re-housed, 59 (66%) had accessed subsidized housing over the study period. Participants who did not access subsidized housing spent, on average, 45% of their income on housing at Time 2, while those who did access subsidized housing spent an average of 31% (t (70) = 2.60, p = .01, two-tailed).

Respondents at Follow-up vs. Non-respondents

Tracking efforts yielded a follow-up rate of 66%. Respondents (n = 101) were compared to non-respondents (n = 53) on scores of variables in the model at Time 1 to determine if there were differences between the two groups. No significant differences were found in terms of age, marital status, level of education, whether participants were born in Canada, or immigration status. Respondents had lived in Ottawa for longer on average than non-respondents, t (152) = -2.28, p < .05, two-tailed. Long time residents may have been more likely to remain in Ottawa, and may have also provided more local contacts than shorter-term residents, making it easier to re-contact them for follow-up.

There were no differences between respondents and non-respondents on any of the predictor variables used in the physical integration model, which included family status, perceived social support, lifetime history of homelessness, level of mental health functioning, and employment history. No comparisons were made for access to subsidized housing as this was measured at Time 2; however, respondents and non-respondents were equally likely to report being on the social housing waiting list at Time 1.

Data Screening

Data were screened for missing values on all predictor variables. SPSS MVA was used to replace missing data for composite measures where at least 80% of items were completed. In cases where less than 80% of items were completed, these cases were excluded from the regression analysis. Missing data were replaced for 4 cases on the employment history variable, and for 5 cases on the social support variable. Results of evaluation of assumptions for the regression analyses led to transformations of selected variables to reduce skewness and improve the normality, linearity, and homoscedasticity of residuals. However, as overall results did not differ from those obtained using untransformed variables, the untransformed variables were retained. With the use of a p < .001 criterion for Mahalanobis distance, no multivariate outliers among the cases were found.

Model of Physical Integration

A logistic regression was conducted to predict housing status at follow-up, and a multiple regression was conducted to predict housing stability at follow-up. An a priori power analysis was conducted using G*Power 3 (Faul, Erdfelder, Lang, and Buchner, 2007). In order to detect a medium effect size (f = 15) for a regression equation with 6 predictor variables with power = .80, a sample size of n = 98 is needed. Therefore, our sample size of N = 101 was considered to have sufficient power to detect a medium effect for relationships between predictors and predicted variables.

Housing Status

A post-hoc, direct logistic regression was conducted using PASW Statistics 18 to examine whether the physical integration model would predict participants' housing status at Time 2. A test of the full model with all six predictors against a constant-only model was statistically reliable, [chi square] (6, N = 101) = 17.49, p < .01, indicating that, as a set, the six predictor variables reliably distinguish between women who were housed for at least 90 days at follow-up and those who were not. The model successfully accounted for a moderate amount of variance in housing status, with Nagelkerke [R.sup.2] = .31.

Table 2 presents the regression coefficients, Wald test statistics, odds ratios and 95% confidence intervals for odds ratios for each of the 6 predictor variables. Of the six predictors, two successfully distinguished women who were re-housed at follow-up from those who were not. These were family status, with women with dependent children being more likely to be housed than women who were not accompanied by children, and access to subsidized housing, with women who had accessed subsidized housing being more likely to be housed than those who did not. A series of post-hoc, hierarchical logistic regression analyses was conducted to test for interactions effects between each of the predictor variables on housing status. No significant interaction effects were found.

Housing Stability

A standard multiple regression analysis was conducted using PASW Statistics 18 to test the ability of the physical integration model to predict housing stability. Overall, the model was statistically significant, F (6, 94) = 3.11, p < .01; however, the six predictor variables combined accounted for only a moderate amount (17%) of the variance in amount of time housed ([R.sup.2] = .17). Table 3 shows semi-partial correlations (sr), unstandardized regression coefficients (B), and standardized regression coefficients ([beta]) for the standard multiple regression. Only the family status variable yielded a significant semi-partial correlation, with being in a family predicting greater housing stability. A series of post-hoc, hierarchical multiple regressions was conducted to test for interaction effects between each of the predictor variables on housing stability. No significant interaction effects were found.

Discussion

This study examined predictors of physical integration in the community for a sample of women who have experienced homelessness. The model of physical integration presented was partly supported; however, the results also suggest that predictors of physical integration for this sample of women in a mid-size Canadian city may differ from those found in past American research. Family status and access to subsidized housing appear to contribute to the physical integration of women who have experienced homelessness. Women with dependent children were more likely than women who were unaccompanied by children both to become re-housed, and to achieve housing stability. The difference between the two groups was substantial; odds ratios indicate that women with dependent children were almost ten times more likely than women unaccompanied by children to become re-housed, and were housed for substantially longer (84% vs. 58% of the follow-up period). Although families are given priority access to subsidized housing, the regression equations account for the unique variance contributed by each independent variable, indicating that the relationship between family status and physical integration exists regardless of whether participants had access to subsidized housing.

There are a number of possible explanations for the impact of family status on physical integration. First, in the larger study from which these data were taken, women with dependent children reported fewer personal disabilities overall than did women who were unaccompanied by children. This includes fewer physical and mental health difficulties as well as less substance abuse (Aubry, Klodawsky and Hay, 2003). This is highlighted by the fact that a substantial proportion of the women unaccompanied by children were, in fact, mothers of children under age 18, but did not have their children in their custody at any time during the study period. This suggests that these women were unable to care for their children for a variety of reasons, which may have included personal disability. Mental health functioning did not predict physical integration in this sample; however, examination of a broader range of personal disability variables may have revealed an association between personal disability and physical integration.

Second, women with dependent children tended to be younger than women unaccompanied by children, and some research has shown that a younger age is associated with a greater chance of becoming re-housed (Caton et al., 2005). A third possible explanation of women with dependent children's greater integration has to do with the motivation associated with having children to care for. q-he demands of parenting might have led some women to remain in housing that they might otherwise have left for a variety of reasons. It is also possible that affordable rental units are more readily available for women with dependent children than for women unaccompanied by children, who may not be able to find safe, well-maintained housing that is also affordable.

While both women unaccompanied by children and women with dependent children paid, on average, about a third of their income on housing, women with dependent children tended to have higher incomes and may have been able to afford better housing than women unaccompanied by children. Women may be more likely to exit homelessness if high quality housing is available, and higher quality housing may, in turn, be conducive to greater housing stability. Finally, services to help women with dependent children access housing may be more intensive, effective, or readily available than services for women unaccompanied by children.

Access to subsidized housing was another important predictor of physical integration, with odds ratios indicating that women who had accessed subsidized housing were almost four times more likely to become re-housed than women who did not access subsidized housing. This is not surprising, as access to subsidized housing has been shown to be a strong and consistent predictor of becoming re-housed and of housing stability in past research (Shinn et al, 1998, Wong and Piliavin 1997, Zlotnick, Robertson and Lahiff 1999). In addition, in the current study, women who had accessed subsidized housing paid a significantly smaller proportion of their income on housing compared to those who did not access subsidized housing. Surprisingly, access to subsidized housing was associated with becoming re-housed, but was not predictive of housing stability. One possible explanation for this has to do with the length of time participants had been housed. Women in this study were housed for, on average, 1.44 years. However, there was tremendous variability in the length of time housed, which varied from 0% to 100% of the follow-up period. There is a long waiting list for subsidized housing in Ottawa. In 2009, there were 10 235 households on the waiting list (Alliance to End Homelessness, 2010). Waiting time for subsidized housing for someone who is homeless may range anywhere from one month to one year in Ottawa (personal communication, Social Housing Registry of Ottawa, June 17, 2008), and may be much longer for those who are not homeless. It may be that once women in the study accessed subsidized housing, they were able to remain housed; however, the long wait for subsidized housing means that some women may have accessed subsidized housing only toward the end of the follow-up period.

The remaining predictor variables, employment history, mental health functioning, perceived social support, and lifetime history of homelessness were not supported by the physical integration model. Employment history was used as a measure of human capital, which has been found to contribute to attaining and retaining housing in past research (Caton et al., 2005, Piliavin et al., 1996). It may be that current employment, not lifetime history of employment, is a better predictor of attaining physical integration. However, only eight participants were working at Time 1. The majority of women in the sample had relatively sparse work histories, and few had more than a high school education. Women experiencing homelessness may have few work skills and be unable to get more than very low-wage jobs, which would not provide the income necessary to exit homelessness to stable housing. For women with young children, the costs associated with childcare may pose additional barriers to finding or attaining work.

Although mental health difficulties have been found to be a barrier to housing stability (Wong and Piliavin 1997), mental health functioning was not a significant predictor of physical integration in this sample. However, no specific mental health diagnoses were examined. It may be that for some women, severe mental health difficulties make attaining and retaining housing challenging. However, for others, these difficulties may actually provide access to services and housing options such as supportive housing that would not otherwise be available. A closer examination of the housing trajectories of participants suffering from mental illness might yield greater insight of the impact of mental health functioning on physical integration following homelessness.

Perceived social support, a measure of disaffiliation, was not related to either becoming re-housed or to achieving housing stability for the women in this sample. Overall, participants reported relatively high levels of satisfaction with social support, indicating that despite their homelessness, the women in this sample felt they had available social support. However, it is possible that for a variety of reasons, participants did not call upon their available social supports. It is also possible that the support systems of study participants consisted of people who themselves did not have the resources to provide the kind of instrumental or material support needed to help participants exit homelessness to stable housing.

Length of time homeless, a measure of acculturation to homelessness, was not predictive of physical integration. This finding was surprising given that past research has found that spending a greater length time homeless over one's lifetime is associated with a diminished probability of becoming and remaining housed (Piliavin et al., 1993, Piliavin et al., 1996, Zlomick, Robertson and Lahiff 1999). While these authors have suggested that acculturation may make it difficult to exit homelessness and put some individuals at risk of falling back into homelessness after obtaining housing, this was not true of this sample. It may be that this finding holds true for men and not women, given that the majority of participants (approximately 75-85%) in these earlier studies were men.

Limitations

A limitation of this study was the sample size. Unfortunately, approximately one-third of participants were lost to attrition at follow-up, resulting in relatively small sample of women. A larger sample would have provided to opportunity to examine a wider variety of predictor variables, thus providing a fuller and more accurate picture of what helps women become physically integrated in their communities following homelessness.

Another limitation may be the sampling strategy that was used. The sample may not be representative of all homeless women in Ottawa. Almost equal groups of women unaccompanied by children and adults with dependent children were recruited for the purposes of this study. However, approximately 25% more women unaccompanied by children than adults in families stayed in Ottawa shelters in 2009 (Alliance to End Homelessness 2010). As well, all the women in this study resided in shelters at Time 1, which excluded women living in other situations such as on the street, in abandoned buildings, or staying temporarily with friends or family. It addition, since service providers aided in recruitment of participants, there is the possibility that there may have been some bias in the sample, most likely toward higher functioning individuals.

A third limitation is that the study relied on self-reported information. Self-report data may be vulnerable to the effects of bias or inaccuracy due to faulty memory, lack of information, or reticence on the part of participants. A related limitation is the use of only quantitative data in this study. However, findings from qualitative data were also gathered and have been reported elsewhere (Klodawsky, Aubry, Nemiroff, Bonetta and Willis, 2007). The qualitative data support the importance of providing subsidized housing, and provide additional insights into barriers and facilitators of becoming rehoused.

Another limitation is the nature of the social and housing services in Ottawa, which may differ from those in other cities. This may limit the generalizability of the findings.

A final limitation of the study is the relatively small proportion of the sample that remained unhoused at follow-up. This is clearly a positive outcome for study participants; however, it is difficult to determine what helped these women to achieve physical integration in their community. Exiting homelessness and achieving housing stability are likely complex processes that may not be easily explained by the relatively limited model presented here.

Conclusions and Implications

Overall, the model of physical integration of women who are homeless, developed from previous American longitudinal research, was only partially supported. The conducted research represents the first longitudinal study examining the exiting from homelessness in a sample of Canadian women. There was some good news; a majority of women were housed at follow-up, and many had retained their housing for a significant period of time. However, achieving housing stability was a challenge for many participants. One fifth reported additional episodes of homelessness during the follow-up period. In addition, rents are prohibitively high for many women who have experienced homelessness, putting them at risk for further episodes of homelessness. Among women who had not succeeded in accessing subsidized housing, nearly half (48%) spent more than 50% of their income on housing.

It is clear that provision of subsidized housing is essential to help women exit homelessness. However, additional support services are necessary in order to help some women who have experienced homelessness to achieve housing stability. Overall, improving the financial resources of women who have experienced homelessness, whether by providing subsidized housing, higher rates of social assistance, or better opportunities for employment would go a long way to helping women who have experienced homelessness to become physically integrated in their communities.

It is also clear from these findings that women unaccompanied by children and women with dependent children have different housing outcomes. Existing services appear to be helping women with dependent children to exit homelessness and remain stably housed. However, women unaccompanied by children experience greater difficulty and may need increased services to attain physical integration. As noted above, women who are unaccompanied by children have poorer physical and mental health than women with dependent children, and are more likely to have substance abuse difficulties (Aubry, Klodawsky and Hay, 2003). For some women, services may be needed to address these difficulties in conjunction with housing services.

In addition, women with dependent children may receive more intensive services than women who are unaccompanied by children. In general, greater efforts appear to be made to end and prevent homelessness for families than for adults who are not accompanied by children. Since these efforts appear to be effective in promoting the physical integration of women with dependent children, it is clear that similar efforts need to be made to help women who are not accompanied by children to become stably housed.

Further research is necessary to understand the role of service providers in helping women who are homeless attain and maintain housing. In particular, evaluations of existing services could help pinpoint which areas of current service provision are helpful and which require improvement. Reviews of best practices in the areas of housing placement and housing loss prevention may provide some potential avenues for the improvement of existing services. Finally, larger longitudinal studies may continue to provide important information on factors that help women who have experienced homelessness to become and remain housed.

Acknowledgments

The researchers gratefully acknowledge the funding for the study provided through the Social Sciences Humanities Research Council, Canada Mortgage Housing Corporation, and the National Homelessness Initiative, Government of Canada administered by the City of Ottawa. Please direct any requests for information about this paper to Tim Aubry, Centre for Research on Educational and Community Services, University of Ottawa, Ottawa, ON, Canada K1N 6N5 or taubry@uottawa.ca

References

Alliance to End Homelessness. 2010. Report Card on Ending Homelessness in Ottawa, Jan-Dec 2009. Ottawa, ON: Author.

Anucha, U. 2003. A study of tenant exits from housing for homeless people: Where do they come from? Why do they leave? Where do they go? Research Highlights, Canada Mortgage and Housing Corporation. http://www. cmhcschl.gc.ca/odpub/pdf/63177.pdf. (accessed November 30, 2009).

Aubry, T. and J. Myner. 1996. Community integration and quality of life: A comparison of persons with psychiatric disabilities in housing programs and community residents who are neighbours. Canadian Journal of Community Mental Health 15: 5-20.

Aubry, T., F. Klodawsky, and E. Hay. 2003. Panel Study on Persons Who Are Homeless in Ottawa: Phase 1 Results. Research report prepared for the City of Ottawa Homelessness Initiatives Team. Centre for Research on Community Services, University of Ottawa.

Aubry, T., F. Klodawsky, E. Hay, R. Nemiroff, and S. Hyman. 2004. Developing a Methodology for Tracking Persons who are Homeless over Time. Research report prepared for the Canada Mortgage and Housing Corporation. Centre for Research on Community Services, University of Ottawa.

Aubry, T., F. Klodawsky, R. Nemiroff, S. Birnie, and C. Bonetta. 2007. The Panel Study on Persons Who Are Homeless in Ottawa: Phase 2 Results. Research report prepared for the City of Ottawa Homelessness Initiatives Team. Centre for Research on Community Services, University of Ottawa.

Bassuk, E.L., J.C. Buckner, L.F. Weinreb, A. Browne, S.S. Bassuk, R. Dawson, and J.N. Perloff. 1997. Homelessness in female-headed families: Childhood and adult risk and protective factors. American Journal of Public Health 87: 241-248.

Breese, J.R. and K.M. Feltey. 1996. Role exit from home to homelessness. Free Inquiry in Creative Sociology 24: 67-76.

Buckner, J.C., E.L. Bassuk, and B.T. Zima. 1993. Mental health issues affecting homeless women: Implications for intervention. American Journal of Orthopsychiatry 63: 385-399.

Burr, M.R. and M.E. Cohen. 1989. Differences among homeless women unaccompanied by children, women with children, and single men. Social Problems 36: 508-524.

Caton, C. L. M., B. Dominguez, B. Schanzer, D.S. Hasin, P.E. Shrout, A. Felix, H. McQuistion, L.A. Opler, and E. Hsu. 2005. Risk factors for long-term homelessness: Findings from a longitudinal study of first-time homeless single adults. American Journal of Public Health 95 (10): 1753-1759.

Clapham, D. 2003. Pathways approaches to homelessness research. Journal of Community and Applied Social Psychology 13:119-127.

Cohen, C.I., M. Ramirez, J. Teresi, M. Gallagher, and J. Sokolovsky. 1997. Predictors of becoming redomiciled among older homeless women. The Gerontologist 37: 67-74.

Farrell, S.J., T. Aubry, F. Klodawsky, P. Jewett, and D. Pettey. 2000. Describing the homeless population of Ottawa-Carleton: Fact sheets of selected findings, http:// www.uottawa.ca/academic/socsci/crcs/homeless/. (accessed March 1, 2010).

Farrington, A. and W.P. Robinson. 1999. Homelessness and strategies of identity maintenance: A participant observation study. Journal of Community and Applied Social Psychology 9: 175-194.

Faul, F., E. Erdfelder, A.G. Lang, and A. Buchner. 2007. G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods 39:175-191.

Fisher, B., M. Hovell, C.R. Hofstetter, and R. Hough. 1995. Risks associated with long-term homelessness among women: Battery, rape, and HIV infection. International Journal of Health Services 25:351-369.

Flynn, R. J. and T.D. Aubry. 1999. Conceptualization and measurement of integration. In R.J. Flynn and R.A. Lemay, (eds). A quarter century of normalization and social role valorization: Evolution and impact (271-303).. Ottawa, ON: University of Ottawa Press.

Gaetz, S. 2010. The struggle to end homelessness in Canada: How we created the crisis, and how we can end it. The Open Health Services and Policy Journal 3: 21-26. http://www.bentham.org/open/tohspj/openaccess2.htm (accessed April 15, 2010)

Grigsby, C., D. Baumann, S.E. Gregorich, and C. Roberts-Gray. 1990. Disaffiliation to entrenchment: A model for understanding homelessness. Journal of Social Issues 46: 141-156.

Guest, A.M. and K.R. Stamm. 1993. Paths of community integration. Sociological Quarterly 34: 581-595.

Gulcur, L., S. Tsemberis, A. Stefancic, and R.M. Greenwood. 2007. Community integration of adults with psychiatric disabilities and histories of homelessness. Community Mental Health Journa143: 211-228.

Hulchanski, J.D., P. Campinski, S.B.Y. Chau, S.W. Hwang, and E. Paradis. 2009. Homelessness: What's in a Word? In J.D. Hulchanski, P. Campinski, S.B.Y. Chau, S.W. Hwang and E. Paradis (Eds.). Finding Home: Policy Options for Addressing Homelessness in Canada (e-Book), Cities Centre Press, University of Toronto. http://homelessness.samhsa.gov/resource/ finding-home-policyoptions-for-addressing-homelessness-in-canada-45761.aspx. (accessed April 1, 2010)

Klodawasky, F., T. Aubry, R. Nemiroff, C. Bonetta, and A. Willis. 2007. What happens over time: Researching homelessness longitudinally. Canadian Journal of Urban Research 16 (1): 93-111.

Piliavin, I., M. Sosin, A.H. Westerfield, and R.L. Matsueda. 1993. The duration of homeless careers: An exploratory study. Social Service Review 67: 576-598.

Piliavin, I., B.R.E. Wright, R.D. Mare, and A.H. Westerfield. 1996. Exits from and returns to homelessness. Social Service Review 70: 33-57.

Roll, C. N., P.A. Toro, and G.L. Ortola. 1999. Characteristics and experiences of homeless adults: A comparison of single men, women unaccompanied by children and women with children. Journal of Community Psychology 27: 189-198.

Sarason, I. G., H.M. Levine, R.B. Basham, and B.R. Sarason. 1983. Assessing social support: The Social Support Questionnaire. Journal of Personality and Social Psychology 44: 127-139.

Shinn, M., J.R. Knickman, and B.C. Weitzman. 1991. Social relationships and vulnerability to becoming homeless among poor families. American Psychologist 46: 1180-1187.

Shinn, M., B.C. Weitzman, D. Stojanovic, J.R. Knickman, L. Jimenez, L.

Duchon, S. James, and D. Kranz. 1998. Predictors of homelessness among families in New York City: From shelter request to housing stability. American Journal of Public Health 88:1651-1657.

Snow, D.A. and L. Anderson. 1987. Identity work among the homeless: The verbal construction and avowal of personal identities. American Journal of Sociology 92: 1336-1371.

Thomas, S. and W. Wolfensberger. 1999. An overview of social role valorization. In A quarter century of normalization and social role valorization: Evolution and impact, eds. R.J. Flynn and R.A. Lemay, 125-159. Ottawa, ON: University of Ottawa Press.

Toro, P.A., J.M. Rabideau, C.W. Bellavia, C.V. Daeschler, D.D. Wall, and D.M. Thomas. 1997. Evaluating an intervention for homeless persons: Results of a field experiment. Journal of Consulting and Clinical Psychology 65: 476-484.

Toro, P.A., S.M. Wolf, C.W. Bellavia, D.M. Thomas, L.L. Rowland, C.V. Daeschler, and P.A. McCaskill. 1999. Obtaining a representative sample of homeless persons: A two-city study. Journal of Community Psychology 27:157-177.

United Nations. 2007, November. United Nations expert on adequate housing calls for immediate attention to tackle national housing crisis in Canada [press release], http://www.unhchr.ch/huricane/huricane.nsf/view01/90995D69CE 8153C3C1257387004F40B5?opendocument. (accessed April 12, 2008).

Ware, J.E., M. Kosinski, and J.E. Dewey. 2002. How to score Version 2 of the SF-36[R] Health Survey. Lincoln, RI: Quality Metric Incorporated.

Wong, Y. I. and I. Piliavin. 1997. A dynamic analysis of homeless-domicile transitions. Social Problems 44: 408-423.

Wong, Y.I., I. Piliavin, and B.R.E. Wright. 1998. Residential transitions among homeless families and homeless single individuals: A comparison study. Journal of Social Service Research 24: 1-27.

Yanos, P.T., S.M. Barrow, and S. Tsemberis. 2004. Community integration in the early phase of housing among homeless persons diagnosed with severe mental Illness: Successes and challenges. Community Mental Health Journal 40: 133-150.

Zlomick, C., M.J. Robertson and M. Lahiff. 1999. Getting off the streets: Economic resources and residential exits from homelessness. Journal of Community Psychology 27: 209-224.

Rebecca Nemiroff

Tim Aubry

Department of Psychology

University of Ottawa

Fran Klodawsky

Department of Geography and Environmental Studies

Carleton University
Table 1: Sample Characteristics

                                           Whole Sample
                                            (N = 101)
Variables                                  M (SD) or %

Housing stability (proportion of time
between interviews in which participants   0.70 (0.35)
were housed at Time 2)

Housing status (% housed 90 days or        88%
more)

Family status (% unaccompanied by          52%
children)

Perceived social support (SSQ-S)           4.70 (1.37)

Lifetime history of homelessness           18.65 (28.73)
(months)

Mental health functioning (SF-36 MCS)      35.60 (12.54)

Employment history (proportion of time
worked since age 16)                       42.14 (31.16)

Access to subsidized housing
(% accessing subsidized housing at         60.04%
any time over study period)

Table 2: Results of Logistic Regression Predicting Housing Status

                                                       95% Confidence
                                                       Interval for
                                                       Odds Ratio
                                 Wald        Odds
                                 Test        Ratio
Variables                B     (z-ratio)  ([e.sup.b])  Lower   Upper

Family status           2.28   3.68 *        9.80      .95     100.34
Employment history       .01   1.03          1.01      .99     1.03
Mental health           -.03   .82            .97      .91     1.04
Perceived social
  support                .10   .10           1.10      .60     2.02
Lifetime history of
  homelessness           .00   .00           1.00      .98     1.02
Access to subsidized
  housing               1.78   4.32 *        5.95      1.10    31.94
Constant                 .80   .19           2.23

* p<.05

Table 3: Results of Multiple Regression Predicting
Housing Stability

Variable                            [beta]     sr     B

Family status                         .34     .29    .23 *
Employment history                   -.01    -.01    .00
Mental health                        -.02    -.02   -.001
Perceived social support              .14     .14    .04
Lifetime history of homelessness     -.03    -.03    .00
Access to subsidized housing          .11     .10    .08

* p<.01.
Gale Copyright: Copyright 2010 Gale, Cengage Learning. All rights reserved.