Printer Friendly

An exploration of alienation and replacement theories of social support in homelessness.

This study conducted a preliminary test of two theories (general alienation and replacement) of support networks in relation to duration of homelessness. Nine hundred respondents experiencing homelessness were selected randomly from shelters and day centers and recruited from city streets. Information obtained from interviews included number and perceived reliability of supports--distinguishing between family supports, friend supports with homes, and friend supports without homes. Duration of homelessness was defined by shorter ([less than or equal to] 12 months) versus longer (> 12 months) length of lifetime homelessness. Support for the general alienation theory included associations between longer-term homelessness and smaller family support networks and inability to count on family and friends. Support for the replacement theory included the lack of differences between the two groups in terms of number of friends and the association between longer-term homelessness and inability to count on friends.

Key words: affiliation; homelessness; social support networks


The body of literature on the social support of People experiencing homelessness has grown during the past several years. A theoretical framework in these investigations, however, surprisingly lacking. Most work seems to be aligned with an affiliation-disaffiliation continuum (for example, Grigsby, Baumann, Gregorich, & Roberts-Gray, 1990; Hertzberg, 1992; Jackson-Wilson & Borgers, 1993; LaGory, Ritchey, & Fitzpatrick, 1991; Stovall & Flaherty, 1994). In this affiliation-disaffiliation model, attachments to oneself, others, and society's institutions theoretically represent anchors that prevent individuals from floating into the disconnectedness associated with vulnerability to homelessness. The purpose of this study was to examine three social support scenarios, based on affiliation-disaffiliation theory, by comparing size and reliability of supports with duration of homelessness.


Thinking about the social relationships of people experiencing homelessness has progressed from earlier ethnographic and qualitative research. It appears that the social networks of individuals experiencing homelessness evolve in a social and cultural context based on the pursuit of survival needs and modified by other perceived needs such as alcohol and other drugs (for example, Anderson, 1978; Liebow, 1967; Snow & Anderson, 1987). In the past two decades, quantitative research has operationalized social support in terms of network size.

Naturalistic studies of the social support of individuals and families experiencing homelessness described relatively small networks (LaGory et al., 1991; Solarz & Bogat, 1990; Stovall & Flaherty, 1994). Subsequent studies comparing people experiencing homelessness with those with low incomes and homes found smaller networks among those experiencing homelessness (Bassuk et al., 1996; Jackson-Wilson & Borgers, 1993; Letiecq, Anderson, & Koblinsky, 1998; Passero, Max, & Zozus, 1991). This difference has not been found universally, however. One study reported no difference between the two groups (Goodman, 1991), and another study described larger networks among those experiencing homelessness (Shinn, Knickman, & Weitzman, 1991). Shinn and colleagues have alleged that people experiencing homelessness "use up" or drain their network resources during their journey into homelessness, accounting for their counterintuitive finding.

More recently, the social networks among particular subgroups of individuals experiencing homelessness have been examined. Among people with mental illness, smaller social networks were found in those experiencing homelessness than in those with homes (Segal, Silverman, & Ternkin, 1997; Wu & Serper, 1999). But these networks have been shown to be similar in size to others experiencing homelessness without mental illness (Segal et al.). Of people experiencing homelessness, smaller networks were found among individuals with alcohol problems than among those without alcohol problems (Sosin & Bruni, 1997). The relationships are not as clear in studies examining individuals using drugs. Among people injecting drugs, larger drug networks were found among those experiencing homelessness compared with those with homes (Latkin, Mandell, Knowlton, Vlahov, & Hawkins, 1998). Women experiencing homelessness who use noninjection drugs have reported larger networks than two other subgroups: those experiencing homelessness who do not use drugs and those experiencing homelessness who use both injection and noninjection drugs (Nyamathi, Flaskerud, Leake, & Chen, 1996).

One complicating factor in interpreting this literature involves the composition of a person's networks. Studies examining differences between formal (that is, service provider) and informal (that is, family, friends, and acquaintances) supports found no difference in the number of formal supports between people experiencing homelessness and those with homes (Goodman, 1991; Letiecq et al., 1998). But subgroups of those experiencing homelessness--such as individuals with mental illness and substance use problems--may have larger numbers of service providers in their networks. Differences have not been examined among individuals experiencing homelessness with and without mental illness, and no difference was found between individuals experiencing homelessness and the use and nonuse of injection drugs (Latkin et al., 1998).

The types of informal supports a person has may play a role as well. Individuals with alcohol and other drug use problems may have networks composed largely of others with substance use problems. If this is the case, having a high number of supports may not be helpful. Nyamathi, Leake, Keenan, and Gelberg (2000) studied the members in the networks of women experiencing homelessness. Although they did not report information strictly on women with alcohol and other drug use problems, they reported that 5 percent of their sample reported support primarily from substance users, and 11 percent of their sample reported support from a mixture of substance users and nonusers. Although these numbers are not high, they may explain, in part, why their noninjection drug users reported larger networks than others.

Other types of negative support (that which is annoying, draining, and hurtful) are possible. Few studies have examined this concept. One study detected no differences in number of negative supports between people experiencing homelessness and those with homes (Goodman, 1991). The findings may be different when examined among people with mental illness and substance use problems.

Individuals experiencing homelessness are likely to derive different resources from relationships with their family members and friends. Compared with people who have low incomes and homes, individuals experiencing homelessness had smaller family networks, but equivalent-sized friendship networks (Wu & Serper, 1999). Among those with mental illness, individuals experiencing homelessness had fewer friends than those with homes (Segal et al., 1997). In contrast, among injection drug users, the experience of homelessness was associated with a larger friendship network (Latkin et al., 1998).

Investigating social networks in terms of size alone underestimates their complexity. Some studies examined the resources individuals report to be in their networks and found that people experiencing homelessness had fewer resources than those with homes (Bassuk et al., 1996; Latkin et al., 1998; Passero et al., 1991). Fewer resources among support network members indicates less ability to provide assistance when needed, thus, representing less protection against becoming and remaining homeless.


Three scenarios of entry into and the course of homelessness are presented in this article. Each scenario, based on affiliation-disaffiliation theory, describes a pathway in relationship to a person's support network.

General Alienation

This first scenario involves a decline in the size and reliability of support networks with entry into and increasing chronicity of experiencing homelessness. The process of entering into and experiencing homelessness may involve alienation from previously existing supports. The reliability of existing relationships may degenerate with increasing time of experiencing homelessness. Evidence for this general alienation theory would be finding a smaller and less reliable support system associated with greater duration of homelessness, operationalized in the following hypothesis: Hypothesis 1--A greater duration of homelessness will be associated with a smaller and less reliable support network.


The second scenario involves replacing the previous support system with a weaker one. It was conceptualized by Hertzberg (1992), who described the selection of new supports to replace the members lost during the "downward spiral into homelessness." The individuals who replace older supports represent new relationships and experience homelessness themselves; therefore, they cannot provide as much support as the individuals they replaced. Thus, although the number of supportive members remains relatively constant, they lack the resources and the reliability of the previous relationships to help the individual obtain a stable home effectively. Grigsby and colleagues (1990) distinguished among family supports, friend supports with homes, and friend supports experiencing homelessness. They described the replacement of friend supports with homes with new friend supports experiencing homelessness as producing less stable, weaker friendships. This downward spiral scenario would suggest the following hypothesis: Hypothesis 2--The duration of homelessness is associated not with lower numbers of support, but with less reliable relationships.

Social Support Unrelated to Homelessness

In the third scenario, social support bears no relationship to the duration of homelessness. This final pattern is a null hypothesis: Hypothesis 3--The duration of homelessness is not associated with size or reliability of the social support network.

Associations predicted by these hypotheses in number and reliability of family and friend relationships among shorter-term versus longer-term homeless people were compared in an epidemiologic sample of men and women experiencing homelessness.



In St. Louis, Missouri, 600 men and 300 women were recruited from overnight shelters and day centers that serve people experiencing homelessness and from locations on the street or other public areas where individuals experiencing homelessness are known to congregate. Respondents were considered experiencing homelessness if they had no stable residence and were living in a public shelter or in an unsheltered location without a personal mailing address, such as on the streets, in a car, in an abandoned building, or in a bus station. People living in marginal housing, such as those doubled up with friends or relatives or living in single-room occupancy facilities, were not included in the sample.

The 14 overnight shelters for men and all but two of the 13 overnight shelters for women, all day centers, and both rehabilitation programs in St. Louis cooperated with the project. Because each night shelter and day center tended to attract a slightly different subpopulation of people experiencing homelessness, sampling was conducted proportionally to the numbers of people in the various programs. The list of shelters and day centers was randomized, and a set of random numbers was generated by computer to select respondents from the daily log of residents occupying beds in the overnight shelters or attending the day centers. This random sampling procedure provided a shelter sample believed to be representative of men and women experiencing homelessness in St. Louis who use shelters and day centers. Interviews were conducted at the various sites at different times of the year.

The majority (70 percent) of the 600 men and all 300 women were recruited from shelters. Interviews were conducted with 195 men and 251 women from overnight shelters, 150 men and 29 women from day centers, 76 men from specialized rehabilitation programs, and 20 women from 24-hour emergency shelters.

The remaining 179 men were recruited from streets, parks, and other public areas. Sampling this "street" population required development of street sampling routes (both walking and driving), each of which consisted of a loop circuit that was systematically covered by interviewers recruiting consecutively encountered individuals experiencing homelessness on these routes. Although plenty of men experiencing homelessness were found on the streets, almost no women experiencing homelessness were available for interview from these locations. More details of the street sampling methods used are provided elsewhere (Smith, North, & Spitznagel, 1991).

Professionally trained interviewers conducted all interviews, and interviews lasted two hours on average. Respondents received $10 for their participation. The interviewing was completed in 12 months. The interview completion rate among eligible respondents was 91 percent for men and 96 percent for women. Before the inception of this study, approval was obtained from the Human Studies Committee of the Washington University School of Medicine, and informed consent was obtained from all respondents before they were interviewed.

The sample was relatively young (M= 33.6 years, SD = 10.6), with ethnic minority individuals predominating (75 percent African American), marital partners lacking (5 percent married), low educational attainment indicated (M = 11.6 years, SD = 2.2), and unemployment prevailing (71 percent). The most prevalent lifetime diagnoses among men were alcohol (63 percent, n = 369 and drug (40 percent, n = 222) use disorders, and among women, major depression (25 percent, n = 74) and drug use disorder (26 percent, n = 75). More details of demographic and diagnostic characteristics of the sample are described elsewhere (Smith et al., 1992, 1993).

Although these data were collected in 1989-1990, the demographic characteristics were not different from those of a sample collected with almost identical methods in 1999-2001 (North et al., 2002). In the newer sample, individuals were slightly older, and more were employed full-rime. Rates of psychiatric diagnoses were higher in the newer sample, but the ranking of the most prevalent diagnoses remained the same. For these reasons, this research team believes we believe the older sample is appropriate for these analyses. This article is the first from this dataset to examine social support grounded in a theoretical context.


Trained lay interviewers administered the Diagnostic Interview Schedule (DIS) for the DSM-III-R (Robins, Helzer, Cottler, & Goldring, 1989), the Homeless Supplement (DIS/HS) (Smith & North, 1989), and several questions on social support. The DIS is a structured interview that asks respondents whether they have ever experienced each symptom of several lifetime diagnoses, including alcohol and drug use disorders, major depression, schizophrenia, and bipolar affective, panic, generalized anxiety, and posttraumatic stress disorders. Psychiatric symptoms were scored positive if they fit criteria for clinical significance and were not explained entirely by physical illness or substance ingestion. This version of the DIS has not been tested for reliability.

The DIS/HS, designed for use in this study, obtained information about respondents' experiences of homelessness, including total number of lifetime years of homelessness. An item-by-item test-retest study found a range of reliability scores (poor to excellent) with this instrument (North et al., 2002). The item--total number of lifetime years of homelessness--yielded a good (0.66) kappa score.

Several social support items, designed for use in this study, elicited information about respondents' relatives and friends. Items regarding family included number of relatives in recent (12-month) contact and how many the respondent could count on. Questions about friends included number of friends, how many were experiencing homelessness, and how many the respondent could count on. These questions have not been tested for reliability.

Independent variables included demographics, substance abuse and dependence, other psychiatric disorders, and length of time homeless. Demographic variables were sex, race (expressed dichotomously as white/not white), years of education (with general equivalency diploma = 12), and history of marriage (expressed dichotomously as never/ever married). Substance use disorder variables were alcohol use disorder and drug use disorders (that is, cannabis, amphetamines, sedatives, prescription medications, cocaine, heroin, opiates, PCP, hallucinogens, inhalants, and other drugs). Other psychiatric disorder variables were major depression, bipolar affective disorder, anxiety disorder, and schizophrenia. Generalized anxiety disorder, panic disorder, and posttraumatic stress disorder were combined into one anxiety variable because analysis of anxiety disorders separately yielded no differences in results. All psychiatric diagnosis variables were expressed dichotomously as not present/present.

Data Analysis

Descriptive data are summarized with percentage rates and means with standard deviations. The variable reflecting continuous number of months of homelessness was markedly skewed and was dichotomized using a 12-month cutoff(Grigsby et al., 1990). The sample was divided dichotomously into "shorter term" (12 or fewer months) and "longer term" (more than 12 months) homeless groups. To detect differences in social support between the shorter- and longer-term groups and to detect relationships among the social support variables, chi-square tests were performed with categorical variables, and t tests with continuous variables. Square roots of dependent variables were used to correct for significant skew found in all. Whereas test statistics are presented for transformed variables for ease of interpretation, means and standard deviations for non-transformed variables are provided.

To predict continuous dependent variables, we performed simple and multivariate linear regression analyses. Independent variables demonstrating significant relationships in simple linear regression equations were entered into multivariate linear regression equations as covariates to control for their effects on the dependent variables. Because age was confounded with length of homelessness, it was not included as a covariate. For prediction of dichotomous dependent variables, we performed logistic regression analyses using the same covariate independent variables, first in simple logistic models and later in multivariate logistic models.


The analysis found differences between the shorter- and longer-term groups experiencing homelessness on the independent variables. Compared with respondents in the group experiencing more than 12 months of homelessness, fewer respondents in the shorter-term group were men [54 percent versus 81 percent; [chi square] (1, N= 883) = 69.83,p [less than or equal to] .0001] and met lifetime diagnostic criteria for alcohol [41 percent versus 60 percent, [chi square] (1, N = 862) = 29.66, p [less than or equal to] .0001] and cannabis use disorders [11 percent versus 17 percent, [chi square] (1, N = 818) = 5.33, p [less than or equal to] .05] and schizophrenia [3 percent versus 9 percent, [chi square] (1, N= 881)= 14.13, p [less than or equal to] .001] . Conversely, more respondents in the shorter-term group were not white [83 percent versus 72 percent, [chi square] (1, N = 883) = 16.61, p [less than or equal to] .0001], and members of this group reported more education (11.86 [less than or equal to] 2.00 years versus 11.26 [less than or equal to 2.24 years, t = 3.80, p [less than or equal to] .001).

More individuals reporting that they could count on their relatives than those who did not also reported being able to count on their friends [85 percent versus 49 percent, [chi square] (1, N = 492) = 75.75, p [less than or equal to] .0001] . These individuals also reported more family supports (6.65 [less than or equal to] 8.24 versus. 3.84 [less than or equal to] 6.89, t = 7.92, p [less than or equal to] .0001) and friend supports with homes (8.11[less than or equal to] 17.70 versus 3.52 [less than or equal to] 6.94, t= 4.18,p [less than or equal to] .0001). More individuals reporting that they could count on their friends than those who did not also reported being able to count on their relatives [69 percent versus 27 percent, [chi square] (1, N = 4 92) = 75.75, p [less than or equal to] .0001]. These individuals also reported more family supports (6.09 [less than or equal to]8.23 versus 3.72 [less than or equal to] 4.69, t = 4.01, p [less than or equal to] .0001) and friend supports with homes (7.08 [less than or equal to] 15.38 versus 3.06 [less than or equal to] 6.78, t = 4.90, p [less than or equal to] .0001). No relationships were detected among the number of family supports, friend supports with homes, and friend supports without homes.

Size of Support Network

Respondents reported contact with a combined mean of 12.02 (SD = 18.08) and median of six relatives (in person or by telephone or letter) and friends during the past 12 months. Only 7 percent of respondents indicated they had none. The shorter-term group reported significantly more family and friend contacts than the longer-term group (Table 1). Less education, diagnosis of cocaine use disorder, and longer-term homelessness were associated with report of fewer contacts with family and/or friends when controlling simultaneously for the effects of diagnosis of schizophrenia (Table 2).

Number of Family Supports. Respondents reported contact with a mean of 5.42 (SD = 8.39) and a median of three relatives during the past 12 months. A minority (17 percent) acknowledged no family contacts. The shorter-term group reported more family contacts than the longer-term group (Table 1). Female gender, race other than white, greater education, and shorter-term homelessness were associated with reports of significantly more family contacts in the past 12 months when controlling simultaneously for the effects of alcohol use disorder (Table 3).

Number of Friend Supports with Homes. Respondents reported a mean of 6.06 (SD = 14.12) and a median of 2 friends with homes. Nearly one-fifth (19 percent) of the sample reported no friendships with people with homes. The shorter- and longer-term groups did not differ on this variable (Table 1). Nonwhite race and less education were associated with report of fewer friends with homes when controlling simultaneously for the effects of sex (Table 4). But, duration of homelessness was not associated with number of friends with homes in the multivariate model.

Number of Friend Supports without Homes. Respondents reported a mean of 2.56 (SD = 9.12) and a median of 0 friends without homes. Almost three-quarters (70 percent) described no friendships with people without homes. The longer-term group identified significantly more friends without homes than the shorter-term group (Table 1). Male gender, white race, and greater education were associated with report of significantly more friends without homes when controlling simultaneously for the effects of cocaine use disorder (Table 5). However, the number of friends without homes was not associated with duration of homelessness in the multivariate model.

Reliability of Support Networks

More than four-fifths of the sample reported being able to count on at least one relative or friend for help. The shorter-term group was more likely to report this ability than the longer-term group (Table 1). Greater education, lack of marital history, and shorter-term homelessness were associated with report of ability to count on a relative or friend for support when controlling simultaneously for the effects of sex, race, and alcohol, cocaine, and hallucinogen use disorders (Table 6).

Counting on Family for Support. More than one-half of the sample reported being able to count on at least one relative for help, especially the shorter-term group (Table 1). Greater education, diagnosis of cocaine use disorder, and shorter-term homelessness predicted report of ability to count on relatives when controlling simultaneously for the effects of race and diagnosis of schizophrenia (Table 7). Diagnoses of alcohol use disorder and major depression were associated with denial of ability to count on relatives.

Counting on Friends for Support. More than two-thirds of the sample reported being able to count on at least one friend for help, especially the shorter-term group (Table 1). Less education, a history of marriage, diagnosis of major depression, and longer-term homelessness were associated with denial of ability to count on friends when controlling simultaneously for the effects of sex, race, and diagnosis of schizophrenia (Table 8).

Relationships of Covariates with Supports

Tables 2 through 8 show that higher numbers of contacts with family and friends and perceived ability to count on them were associated consistently with significantly more years of education. Being not white--being overwhelmingly African American in this sample--was associated with significantly more family contacts and significantly fewer friend contacts. Other significant covariates demonstrated no consistent pattern, indicating their statistical significance in data analyses was most likely due to confounding with other variables or Type I errors.


In the multivariate models, longer-term homelessness was associated significantly with smaller family support networks, but not with size of friendship networks. It was also associated with inability to count on family and on friends, reflecting a lesser degree of perceived reliability of the support network. The evidence--the differences between the two groups in terms of family--favors the alienation hypothesis. However, the data demonstrating lack of differences in number of friends, coupled with the longer-term homelessness group reporting less ability to count on their friends, favors the replacement hypothesis. There may be a progression of change in network size (that is, loss of family relationships) and perceived reliability of network support (that is, loss of reliable friends) during homelessness. Such a hypothesis would need to be tested with longitudinal data because a plausible rival hypothesis might be that those in the longer-term group never had any family contacts.

The conventional wisdom suggesting that people experiencing homelessness associate primarily with other people experiencing homelessness is challenged by these findings. Not only did the respondents report more friends with homes than without (on average six friends versus three friends), but one-half of the sample identified no friends without homes. These results suggest that the friendship networks individuals have before becoming homeless and when they cease being homeless may include many individuals living below the poverty line or hovering around it, many of them also cycling in and out of homelessness. This hypothesis needs to be tested with longitudinal data.

The data do not support the popular notion that people experiencing homelessness are socially isolated. Respondents in this study reported substantial numbers (average = 12) of family and friend contacts; only 7 percent reported none. These findings rival those of other studies that found that individuals experiencing homelessness have an average of five to six supports (Passero et al., 1991; Solarz & Bogat, 1990). This difference might be a function of this study's large sample size and complex, random sampling methods. In this dataset, the majority of respondents reported that they could also count on family (80 percent) or friends (70 percent) for support. According to the literature, people experiencing homelessness name, on average, between two and four supports they can count on in times of need (Letiecq et al., 1998). Respondents reported feeling "mixed" to "mostly satisfied" with their support networks (Solarz & Bogat). Because so many respondents in this study reported no friends without homes, it seems that people experiencing homelessness count largely on their friends with homes for assistance.

No mental health diagnoses were associated with number of family or friend supports, and diagnosis of major depression was the only mental illness associated with perceived reliability of the network (both for family and for friends). Individuals experiencing major depression may experience cognitive distortions that cause them to view their networks as less reliable than others. The other null findings are supported by Segal and colleagues' (1997) work.

Diagnosis of substance use disorders demonstrated few relationships with social support. Alcohol use disorder was associated with inability, and cocaine use disorder with greater ability, to count on family. Both of these drugs can cause delusions, distorting a person's perception of reality. More likely, however, because the antisocial behaviors associated with cocaine use often drive away family members, the relationship with cocaine use disorder is a Type I error. Lower numbers of total family and friend supports were associated with cocaine use disorder. Again, the association of cocaine use with antisocial behaviors explains the finding. Those with heroin use disorder reported fewer friend supports without homes. This may be because heroin use is often a solitary activity. If this were true, this difference would also be found in number of friend supports with homes, but that is not the case; therefore, it is likely that this finding is a Type I error. This study did not find alcohol or drug (commonly injected) use disorders to be associated with network size as found in other studies (Latkin et al., 1998; Nyamathi et al., 1996; Sosin & Bruni, 1997). This difference may have been found because this study's sample was not recruited specifically for people with alcohol or other drug use disorders. The overrepresentation of substance abuse in individuals experiencing homelessness suggests that it is probably prevalent among their friend supports with and without homes. It might be, then, that the problems of substance abuse may be entrenched long before homelessness occurs and represent a larger problem; therefore, several differences may not be seen between the groups. This, too, needs to be examined with longitudinal data.

The model statistics for number of family and friend supports and reliability of family and friend supports were poor. Even though all control variables demonstrating significant relationships with each social support variable in simple equations were entered into the multivariate equations, it appears that other important contributing variables were excluded. Other variables might be family history, social skills, characteristics of supports, and time to invest in developing relationships.


This exploratory study has several methodological limitations. Duration of homelessness was operationalized in terms of shorter-term versus longer-term homelessness, recognizing the potential limitations addressed in this section. Data collected were cross-sectional and retrospective, without longitudinal information. This precludes the study's ability to determine causality in the relationships found. Only homeless people themselves reported information obtained for this study, because collateral information from other members of the support network or outside observers was not available to determine the reliability of the information given. Dichotomizing the sample by shorter- and longer-term homelessness at the one-year mark was based on earlier research and was needed for the skewed distribution on homelessness duration. Although this dichotomization on chronicity of homelessness was validated by observation of significant relationships in the data, these cross-sectional data cannot provide longitudinal insight into progress over time or potential change points.

Some key variables lacked specificity. The family contact variable was limited by its lack of accounting for family size that was likely to be confounded with the number of family members in recent contact. The definition of friend was left to the respondents, rather than being operationally defined in behavioral terms. Finally, lack of psychiatric information beyond dichotomous definitions of psychiatric diagnoses limited the study's ability to examine the role played by severity and length of illness on social support.

Service Implications

The findings from this study have implications for programs, policies, and practices with people experiencing homelessness. The many significant differences in social support found between shorter- and longer-term groups suggest that interventions based on supports may require different loci. For the shorter-term group, family supports need to be re-established, and family members should be included in treatment planning and coordination of care. This may be particularly true for women and people of color, for whom family appears to play a particularly important role (according to these data). Families may represent a large untapped resource in interventions to help people exit homelessness.

For the longer-term group, services may need to focus on friendship networks, perhaps developing positive networks in service systems where support can be provided to the entire system. This may be particularly important for men, white people, and people with substance use disorders (according to these data). The potential for system-based social support groups has been documented elsewhere (Pollio, 1999).

Future Research

Two factors are vital for developing a more sophisticated understanding of the relationship between homelessness and social support. First, longitudinal research is needed to explore the relationship between social support and course of homelessness over time. Following people experiencing homelessness over time and examining changes in their social support relationships could provide a more accurate picture of the role social support plays in entering, experiencing an episode of, and recovering from homelessness. Longitudinal research can move the evidence for the theoretical model from inferences developed from associations to tests of causal relationships. Second, examination of the constructs in greater depth is needed, providing more precise measurements of social support, duration of homelessness, and demographic and diagnostic covariates to study the associations and causal relations and test more refined hypotheses. Finally, although the role that social support plays in the course of homelessness has not been fully specified, findings from this and earlier research suggest potential for developing interventions tailored to gaining access to appropriate social supports as part of the treatment of homelessness. Intervention research provides new methods of treatment to this difficult-to-treat population, adding to the theoretical knowledge as assumptions are tested prospectively through their incorporation in treatment models.

Original manuscript received July 25, 2001

Final revision received December 12, 2002

Accepted December 20, 2002

This research was supported by the National Institute on Alcohol Abuse and Alcoholism, grant number AA007549 and the National Institute of Drug Abuse, grant number DA10713 and predoctoral training grant number DA07313. An earlier version of this article was presented at the 46th Annual Program Meeting, Council on Social Work Education, February 29, 2000, New York. The authors thank the anonymous reviewers for their extensive feedback on the earlier drafts of this article.


Anderson, E. (1978). A place on the corner. Chicago: University of Chicago Press.

Bassuk, E. L., Weinreb, L. F., Buckner, J. C., Browne, A., Salomon, A., & Bassuk, S. S. (1996). The characteristics and needs of sheltered homeless and low-income housed mothers. JAMA, 276, 640-646.

Goodman, L. A. (1991). The relationship between social support and family homelessness: A comparison study of homelessness and housed mothers. Journal of Community Psychology, 19, 321-332.

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

Hertzberg, E. L. (1992). The homeless in the United States: Conditions, typology, and interventions. International Social Work, 35, 149-161.

Jackson-Wilson, A. G., & Borgers, S. B. (1993). Disaffiliation revisited: A comparison of homeless and nonhomeless women's perceptions of family of origin and social supports. Sex Roles, 28, 361-377.

LaGory, M., Ritchey, F., & Fitzpatrick, K. (1991). Homelessness and affiliation. Sociological Quarterly, 32, 201-218.

Latkin, C. A., Mandell, W., Knowlton, A. R., Vlahov, D., & Hawkins, W. (1998). Personal network correlates and predictors of homelessness for injection drug users in Baltimore, Maryland. Journal of Social Distress and the Homeless, 7, 263-278.

Letiecq, B. L., Anderson, E. A., & Koblinsky, S. A. (1998). Social support of homeless and housed mothers: A comparison of temporary and permanent housing arrangements. Family Relations, 47, 415-421.

Liebow, E. (1967). Tally's corner. Boston: Little, Brown.

North, C. S., Eyrich, K. M., Pollio, D. E., Foster, D. A. Cottler, L. B., & Spitznagel, E. L. (2002). The Homeless Supplement to the Diagnostic Interview Schedule: Test-retest analyses. Manuscript submitted for publication.

Nyamathi, A., Flaskerud, J., Leake, B., & Chen, S. (1996). Impoverished women at risk for AIDS: Social support variables. Journal of Psychosocial Nursing, 34, 31-39.

Nyamathi, A., Leake, B., Keenan, C., & Gelberg, L. (2000). Type of social support among homeless women: Its impact on psychosocial resources, health and health behaviors, and use of health services. Nursing Research, 49, 318-326.

Passero, J. M., Max, M., & Zozus, R. T., Jr. (1991). Social network utilization as related to family history among the homeless. Journal of Community Psychology, 19, 70-78.

Pollio, D. E. (1999). Group membership as a predictor of service use related behavior for persons "on the streets." Research on Social Work Practice, 9, 575-592.

Robins, L. N., Helzer, J. E., Cottler, L. B., & Goldring, E. (1989). NIMH Diagnostic Interview Schedule (Version III rev.). St. Louis: Washington University.

Segal, S. P., Silverman, C., & Temkin, T. (1997). Social networks and psychological disability among housed and homeless users of self-help agencies. Social Work in Health Care, 25(3), 49-61.

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

Smith, E. M., & North, C. S. (1989). The Diagnostic Interview Schedule Homeless Supplement. St. Louis: Washington University.

Smith, E. M., North, C. S., & Spitznagel, E. L. (1991). Are hard-to-interview street dwellers needed in assessing psychiatric disorders in homeless men? International Journal of Methods in Psychiatric Research, 1, 69-78.

Smith, E. M., North, C. S., & Spitznagel, E. L. (1992). A systematic study of mental illness, substance abuse, and treatment in 600 homeless men. Annals of Clinical Psychiatry, 4, 111-120.

Smith, E. M., North, C. S., & Spitznagel, E. L. (1993). Alcohol, drugs, and psychiatric comorbidity among homeless women: An epidemiologic study. Journal of Clinical Psychiatry, 54, 82-87.

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

Solarz, A., & Bogat, G. A. (1990). When social support fails: The homeless. Journal of Community Psychology, 18, 79-96.

Sosin, M. R., & Bruni, M. (1997). Homelessness and vulnerability among adults with and without alcohol problems. Substance Use and Misuse, 32, 939-968.

Stovall, J., & Flaherty, J. (1994). Homeless women, disaffiliation, and social agencies. International Journal of Social Psychiatry, 40, 135-140.

Wu, T., & Serper, M. R. (1999). Social support and psychopathology in homeless patients presenting for emergency psychiatric treatment. Journal of Clinical Psychology, 55, 1127-1133.

Karin M. Eyrich, PhD, MPE, LCSW, is NIDA postdoctoral fellow, University of Pennsylvania, School of Medicine, Department of Psychiatry and Treatment Research Institute, 600 Public Ledger Building, 150 South Independence Mall West, Philadelphia, PA 19106; e-mail: David E. Pollio, Phi), LCSW, is associate professor, George Warren Brown School of Social Work, and Carol S. North, MD, MPE, is professor, Department of Psychiatry, School of Medicine, Washington University, St. Louis.
TABLE 1--Univariate and Bivariate Statistics of Social Support
Variables and Shorter- and Longer-Term Homeless Groups in St. Louis

 Univariate Statistic by
 Length of Homelessness
 Complete Sample's
 Univariate Statistic Shorter-term

Dependent Variable M SD M SD

Total no. of family and 12.02 [+ or -] 18.08 12.54 [+ or -] 16.74
 friend contacts (a)
No. family contacts (a) 5.42 [+ or -] 8.39 6.77 [+ or -] 9.30
No. friends with 6.06 [+ or -] 14.12 5.73 [+ or -] 13.91
 homes (a)
No. friends without 2.56 [+ or -] 9.12 1.94 [+ or -] 7.94
 homes (a)
Ability to count on 82 88
 family or friends (%)
Ability to count 53 61
 on family (%)
Ability to count 69 76
 on friends (%)

 Univariate Statistic by
 Length of Homelessness


Dependent Variable M SD [chi square] or
 1 ratio

Total no. of family and 10.85 [+ or -] 18.86 t = 3.04 **
 friend contacts (a)
No. family contacts (a) 3.66 [+ or -] 6.69 t = 7.36 ***
No. friends with 5.99 [+ or -] 13.07 t = 0.07
 homes (a)
No. friends without 3.31 [+ or -] 10.50 t = 2.21 *
 homes (a)
Ability to count on 75 [chi square]
 family or friends (%) = 17.02 ***
Ability to count 43 [chi square]
 on family (%) = 21.80 ***
Ability to count 61 [chi square]
 on friends (%) = 13.46 ***

(a) Significance testing was conducted with a square root
transformation * p [less than or equal to] .05.

** p [less than or equal to] .01. *** p [less than or equal to] .001.

TABLE 2--Multivariate Regression Model Predicting
Total Number of Family and Friend Contacts among
Homeless Men and Women in St. Louis

Variable Beta SE T-Ratio

Years of education 0.08 0.03 2.40 *
Cocaine use disorder -0.45 0.17 -2.70 **
Schizophrenia -0.53 0.34 -1.57
Shorter-term homeless 0.35 0.14 2.50 *

NOTES: Analyses conducted with square root transformation.

[R.sup.2] = 0.03, F(4) = 6.26, p < .0001.

* p [less than or equal to] 0.05. ** p [less than or equal to] 0.01.

TABLE 3--Multivariate Linear Regression Model
Predicting Number of Family Contacts for Homeless Men
and Women in St. Louis

Variable Beta SE T-Ratio

Female 0.44 0.11 3.93 ***
Not white 0.48 0.12 4.05 ***
Years of education 0.07 0.02 3.22 **
Alcohol use disorder -0.11 0.10 -1.03
Shorter-term homeless 0.46 0.10 4.55 ***

NOTES: Analyses conducted with square root transformation.

[R.sub.2] = 0.13,F(5) = 21.79, p < .0001.

** p [less than or equal to] .01. *** p [less than or equal to] .001.

TABLE 4--Multivariate Regression Model Predicting
Number of Friend Contacts with Homes for Homeless
Men and Women in St. Louis

Variable Beta SE T-Ratio

Male 0.34 0.17 1.94
White 0.47 0.19 2.40 *
Years of education 0.13 0.04 3.39 ***

NOTES: Analyses conducted with square root transformation.

[R.sup.2] = 0.05, F(3) = 8.09, p < .0001.

* p [less than or equal] .05. p [less than or equal to] .001.

TABLE 5--Multivariate Regression Model Predicting
Number of Friend Contacts without Homes for Homeless
Men and Women in St. Louis

Variable Beta SE T-Ratio

Male 0.32 0.12 2.68 **
White 0.59 0.14 4.27 ***
Years of education 0.06 0.03 2.28 *
Cocaine use disorder -0.19 0.14 -1.37
Heroin use disorder -0.48 0.24 -2.03 *
Longer-term homeless 0.05 0.11 0.42

NOTES: Analyses conducted with square root transformation.

[R.sup.2] = 0.07, F(6) = 8.26, p < .0001.

* p [less than or equal to] .05. ** p [less than or equal to] .01.
*** p [less than or equal to] .001.

TABLE 6--Logistic Regression Model Predicting Ability to
Count on Family or Friends among Homeless Men and
Women in St. Louis

 Wald Odds
Variable Beta SE [chi square] Ratio

Female 0.13 0.29 0.20 0.88
Not white 0.44 0.27 2.71 1.55
Years of education 0.11 0.05 4.85 * 1.12
Never married 0.49 0.23 4.49 * 0.61
Alcohol use disorder -0.17 0.26 0.41 0.85
Cocaine use disorder 0.64 0.34 3.69 1.90
Hallucinogen use disorder -1.58 0.83 3.65 0.21
Shorter-term homeless 0.70 0.24 8.43 ** 0.50

NOTE: Likelihood Ratio= 41.62, df = 8, p <.0001.

* p [less than or equal to] .05. ** p [less than or equal to] .01.

TABLE 7--Logistic Regression Model Predicting Ability to Count on
Family for Homeless Men and Women in St. Louis

 Wald Odds
Variable Beta SE [chi square] Ratio

Not white 0.28 0.20 2.03 1.33
Years of education 0.07 0.04 4.04 * 1.08
Alcohol use disorder -0.37 0.17 4.92 * 0.69
Cocaine use disorder 0.43 0.20 4.52 * 1.54
Major depression -0.57 0.20 8.07 ** 0.57
Schizophrenia -0.32 0.40 0.65 0.73
Shorter-term homeless 0.57 0.16 12.64 *** 0.57

NOTE: Likelihood Ratio= 50.39, df = 7, p < .0001.

* p [less than or equal to] .05. ** p [less than or equal to] .01.
*** p [less than or equal to] .001.

TABLE 8--Logistic Regression Model Predicting Ability to
Count on Friends among Homeless Men and Women in

St. Louis Wald Odds
Variable Beta SE [chi square] Ratio

Female 0.29 0.24 1.38 0.75
Not white 0.37 0.24 2.44 1.45
Years of education 0.17 0.05 10.96 *** 1.18
Never married 0.45 0.22 4.30 * 0.64
Major depression -0.57 0.26 4.89 * 0.57
Schizophrenia -0.95 0.53 3.23 0.39
Shorter-term homeless 0.55 0.22 6.12 * 0.58

NOTE: Likelihood Ratio= 44.10, df= 7, p < .0001.

* p [less than or equal to] .05. *** p [less than or equal to] .001.
COPYRIGHT 2003 National Association of Social Workers
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2003 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Eyrich, Karin M.; Pollio, David E.; North, Carol S.
Publication:Social Work Research
Article Type:Report
Date:Dec 1, 2003
Previous Article:The supportive community: help seeking and service use among elderly people in Jerusalem.
Next Article:Exploring the relevance of moral development as a treatment issue in batterer intervention.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters