Homelessness
chales28
2.pdf
Social Service Review (September 2008). � 2008 by The University of Chicago. All rights reserved. 0037-7961/2008/8203-0005$10.00
Homelessness among At-Risk Families with Children in Twenty American Cities
Angela R. Fertig University of Georgia
David A. Reingold Indiana University–Bloomington
This article uses data from the Fragile Families and Child Wellbeing study to explore the characteristics and determinants of homelessness among families with children. These unique data permit the examination of a large set of individual-, household-, and city- level risk factors that may influence homelessness. Results suggest that homelessness is strongly linked to informal and institutional social support. It is only modestly associated with local housing and labor market conditions. These results suggest that the greatest potential for reducing family homelessness lies in interventions, such as low-income hous- ing assistance, that are designed to strengthen informal and institutional social support among low-income mothers. Policies designed to alter local housing and labor market conditions are unlikely to reduce substantially the risk of this pressing social problem.
Scholarly research over the past 25 years has firmly documented the emergence and persistence of family homelessness in the United States. Recent estimates from the U.S. Department of Housing and Urban Development (HUD 2007) suggest that there were approximately 754,147 sheltered and unsheltered homeless persons in the United States in January 2005. Between one-third and one-half of homeless persons are part of families that include children. An estimated 215,000 beds in homeless shelters are dedicated to serving these families (HUD 2007). These numbers exclude those families that are doubled up with friends and family.
The notion of the homeless household, usually a mother and her children, represents a departure from the stereotypical image of home-
486 Social Service Review
less individuals as skid row residents who are predominately single, work- ing-age males. Many homeless adults (60 percent of homeless women and 41 percent of homeless men) have minor children, but only 28 percent of homeless parents live with their children. Among those chil- dren living with a homeless parent, 80 percent are under 11 years of age (Burt et al. 1999; Burt 2001b).
Homeless families can be found living in cars, abandoned buildings, and homeless shelters within virtually every American city. Some families are able to avoid this type of severe material hardship by doubling up with friends and family. Many theories have attempted to explain the growth of family homelessness (see Shlay and Rossi [1992] and Jencks [1994] for detailed historic trends and additional background infor- mation on homelessness in the United States). The growth in single- parent families and the decline of marriage may have left women and children more vulnerable to numerous economic hardships, including homelessness, now than in the past ( Jencks 1994). In addition, Miller McPherson, Lynn Smith-Lovin, and Matthew Brashears (2006) argue that there has been a general decline in the availability of close personal relationships from which informal social support can be drawn; this decline may make people increasingly vulnerable during times of per- sonal and economic crisis. Moreover, the vulnerability may be worsened by rising housing prices and limited wage growth among unskilled work- ers (Quigley, Raphael, and Smolensky 2001; Lee, Price-Spratlen, and Kanan 2003). There is substantial evidence to suggest that circumstances and events, if considered together, make families a permanent feature of the U.S. homeless population.
Various studies attempt to identify the factors that tend to place fam- ilies at risk of becoming homeless or of doubling up with friends and relatives (Bassuk and Rosenberg 1988; Benda 1990; Wood et al. 1990; Goodman 1991; Bassuk et al. 1998; Bassuk and Geller 2006; Stainbrook and Hornik 2006; Lehmann et al. 2007; Tischler, Rademeyer, and Vos- tanis 2007; Tischler and Vostanis 2007). Many of the studies examine factors likely to also be associated with the rise of family homelessness. These factors include access to economic resources, availability of non- economic social support, family structure, family size, educational at- tainment, local housing market conditions, exposure to domestic vio- lence, history of mental illness, and history of drug abuse. Unfortunately, scientific studies of family homelessness fail to consider all of these likely risk factors. Much of the research can be divided into two categories: studies that present family homelessness as the product of individual characteristics (e.g., Sosin 1989; Wright et al. 1998) and those that pre- sent it as a product of community (or structural) circumstances (e.g., Bohanon 1991; Elliott and Krivo 1991; Glomm and John 2002; Lee et al. 2003). Few studies examine both sets of factors together (Shlay and
Urban Family Homelessness 487
Rossi 1992; Main 1998; Burt 2001a). This division occurs because of the types of data available to study homelessness.
Individual-level data usually come from a single community (or city). The study designs limit what can be said about the effects of variation in local housing stock and labor market conditions on homelessness. Such data also limit researchers’ ability to assess the influence of climate, the supply of shelter beds, and the presence of local laws designed to discourage homelessness. Likewise, community-level data do not allow an examination of individual risk factors. Study limitations are fre- quently compounded by the tendency to collect data from individuals who are homeless at the time of the interview or who have sought assistance from homeless shelters, without collecting comparable infor- mation from individuals who have not been homeless (Phelan and Link 1999). Studies that incorporate comparison groups have no way to select the best comparison group (Wong, Piliavin, and Wright 1998; Dworsky and Piliavin 2000). As a result, much of this literature is unable to provide reliable and systematic information on why some at-risk families become homeless and others do not.
The current study overcomes many of these limitations by examining data from a sample of households, some of which have been homeless or doubled up, across 20 cities that vary by characteristics that may be responsible for creating conditions that can lead to homelessness. Be- cause these data are longitudinal, it is possible to observe the presence of particular risk factors and to measure their influence on homelessness or doubling up over time. The analyses are organized around three research questions: (1) What are the characteristics of homeless and doubled-up families, and how do such characteristics compare with those of a similar subgroup that does not experience a homeless spell during the study period? (2) What factors seem to inoculate at-risk families from experiencing homelessness or doubling up? (3) What are the respective effects of individual and community factors in explaining a family’s exposure to a homeless spell or to doubling up? Overall, the data provide a unique opportunity to understand the relative impor- tance of such factors in exposure to unstable housing arrangements.
Explanations for Family Homelessness
In the homelessness literature, explanations that associate homelessness with individual-level factors focus on characteristics specific to individ- uals or to individual households. In these studies, the choice of com- parison group is not clear-cut but may have a large influence on the findings. Associated individual characteristics include physical health, mental health, substance abuse, addiction, domestic violence, single motherhood, welfare receipt, and educational attainment (Sosin 1989;
488 Social Service Review
Wright et al. 1998). In particular, family homelessness is found to be closely associated with female-headed households, unwed child rearing, and the economic hardships of single mothers (Weitzman 1989; Bassuk et al. 1996). Several researchers find that domestic violence, drug use, and mental illness increase the risk of becoming homeless (Bassuk and Rosenberg 1988; Benda 1990; Wood et al. 1990; Goodman 1991; Bassuk et al. 1998). In contrast, other researchers find that exposure to domestic violence and drug use do not differentiate those who become homeless from those who do not (Bassuk et al. 1997).
There is some debate about whether homeless families lack networks of social support (i.e., relatives and friends to which families can turn for help). Some find that homeless families have little support (Letiecq, Anderson, and Koblinsky 1998). Others argue that these families actually have high levels of contact with their network but exhaust its resources before becoming homeless (Shinn, Knickman, and Weitzman 1991; Toohey, Shinn, and Weitzman 2004). Finally, low family income and low labor force participation are found to be determinants of unstable hous- ing situations that lead to homelessness (Wood et al. 1990; Shinn et al. 1998).
Community (or structural) explanations focus on characteristics be- yond the individual. For example, these characteristics include lack of affordable housing, slack labor markets, welfare reform, the availability of public housing, and access to homeless shelters. Studies of aggregate levels of homelessness in metropolitan areas show that lack of affordable housing is positively associated with rates of homelessness (Quigley et al. 2001; Lee et al. 2003). So too, public housing and other low-income housing subsidies are shown to protect families from experiencing mul- tiple homelessness spells (Bassuk et al. 1997; Wong, Culhane, and Kuhn 1997), although these subsidies are not well targeted to the homeless (Early 1998, 2004). Welfare and other cash benefit programs are also found to have a protective effect against family homelessness (Salomon, Bassuk, and Brooks 1996).
Two important federally funded low-income programs have under- gone substantial changes over the past 15 years, providing researchers an opportunity to study the effect of the efforts on homelessness. The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (U.S. Public Law 104-193) introduced time limits and work re- quirements for receiving welfare. Effects of the law have been studied extensively. Research finds that the implementation of welfare reform in some states resulted in homelessness among welfare leavers; however, the numbers are extremely small, suggesting that the protective effect of welfare may be small (Loprest 1999; Institute for Family and Social Responsibility 2000; Bloom, Farrell, and Fink 2002).
The second significant policy change involves the implementation of the Housing and Opportunities for People Everywhere (HOPE) VI Pro-
Urban Family Homelessness 489
gram (U.S. Public Law 102-389 [1992]) in 1993 and the Quality Housing and Work Responsibility Act of 1998 (Title V of U.S. Public Law 105- 276). These policies attempt to address substantial deterioration of the public housing stock while promoting mixed-income replacement hous- ing. Moreover, the 1998 act gives local housing authorities increased flexibility in the operation of programs, allowing them to favor higher- income housing applicants over very poor applicants. Over the past decade, hundreds of thousands of run-down public housing units have been demolished, but the number of units torn down is far greater than the number of replacement units (Imbroscio 2008).
The Urban Institute is conducting the only national-level effort to track families affected by these changes in low-income housing policy. The study finds that such changes have few, if any, effects on home- lessness (Urban Institute 2004). The study sample is limited, however, to the population of public housing residents who held formal lease agreements with their local housing authority. Another study examines Chicago’s public housing transformation efforts (Venkatesh et al. 2004). Its sample includes lessees as well as individuals who live in public hous- ing illegally as squatters (i.e., those occupying vacant or uninhabitable units or illegally doubled up with lessees). The findings suggest that these structural policy changes may have a dramatic effect on the prob- lem of family homelessness. This study finds that 13 percent of the squatter population is homeless 1 year after building closure (Venkatesh et al. 2004). Thus, public housing may have an important protective effect, particularly for illegal residents of public housing.
In addition to these individual and environmental correlates of home- lessness, the focus of existing studies also falls on the several specific causal pathways that may precipitate severe economic hardship (in- cluding homelessness) among families. Studies vary in their emphasis on the lack of material resources and human capital, frequently em- phasizing the factors within the context of an unexpected crisis linked to domestic violence, mental illness, or drug abuse. Studies also consider pathways to homelessness through single parenthood in low-income households, through the absence of social support (i.e., from relatives and friends), and through the lack of affordable housing in many large metropolitan areas. Each factor is discussed below in turn.
Lack of Material Resources
Research frequently assumes that a primary explanation for homeless- ness is a lack of material resources that leads to the loss of a home (e.g., Shinn et al. 2007); the lack of material resources leading to the loss of a home is often linked to limited human capital and the decline in wage returns among unskilled workers (Ma, Gee, and Kushel 2008). These characteristics alone, however, do not explain why most poor
490 Social Service Review
families do not become homeless or why other poor families that do experience homelessness are able to remove themselves from this sit- uation. Accordingly, one assumption is that homelessness may occur due to the combination of poverty and a sudden crisis brought on by domestic violence, health-related problems, or drug abuse (Culhane et al. 2007).
Family Structure and Social Support
Changes in family structure, such as the increase in single-parent families and the decline of marriage rates, may lead to homelessness by making families more vulnerable to changing economic conditions ( Jencks 1994). Further, as families become increasingly dependent on single wage earners and informal social support (from friends and relatives) during a crisis, low-income housing subsidies, welfare, and other social welfare benefits may gain in importance in forestalling episodes of homelessness.
Local Housing Market Conditions
The lack of affordable housing in many large metropolitan areas and the rising share of household income spent on housing costs may put low-income families at risk of being unable to own or rent a housing unit (Quigley et al. 2001). These local housing market conditions may price some families out of stable housing and may force others to double up with friends or relatives to avoid a homeless spell. Moreover, the availability of shelter beds in a particular city and other conditions may affect the chances that a family will become homeless. Such other con- ditions may include climate and the extent to which a local ordinance criminalizes being homeless.
The empirical family homelessness literature unfolds in a piecemeal fashion and is frequently unable to encompass fully the theoretical par- adigm necessary to recognize both individual and structural factors that generate risk. Because of this fragmentation in the homelessness liter- ature, it is known that individual factors are associated with family home- lessness but not whether these relationships hold if analyses control for structural conditions. Similarly, it is known that structural conditions matter but not whether these relationships remain robust if estimates control for individual protective factors. As a result, the extant literature on family homelessness cannot determine which risk factors matter most. This limits the ability to target interventions on the most important factors that affect the onset of a homelessness episode among at-risk families.
Urban Family Homelessness 491
A Comprehensive Empirical Inquiry
Data and Measures
This article examines individual and structural predictors of homeless- ness by analyzing data from the Fragile Families and Child Wellbeing (FFCW) study. The FFCW follows a birth cohort of nearly 5,000 children born between 1998 and 2000 in 20 U.S. cities with populations greater than 200,000.1 The FFCW study population includes an oversample of births to unwed parents. A stratified random sampling strategy was used to select among large U.S. cities grouped according to their policy en- vironments and labor market conditions. Interviews are conducted with the mother and (separately) the father of the cohort member. The first, or baseline, interview occurs in the hospital after the birth of the cohort member. Data are also available from two follow-up interviews with moth- ers and fathers. One follow-up interview takes place approximately 1 year after the birth, and another occurs 3 years after the birth.2 All follow-up interviews were conducted over the telephone.
The FFCW data are well suited to the goals of this analysis because they capture a population of at-risk families and collect information on whether a respondent was homeless or doubled up. In addition, the data provide a large amount of sociodemographic and life-history in- formation on each respondent. One limitation for this analysis is that the sample is restricted to families with young children and so is not representative of all forms of family homelessness.
Homelessness is defined using responses to two survey items. A family is considered homeless in the period leading up to the interview if the mother indicates (1) that she lives in temporary housing, in a group shelter, or on the street at the time of the interview or (2) that, in the 12 months prior to the interview, she stayed in a shelter, an abandoned building, an automobile, or any other place not meant for regular hous- ing, even for 1 night. In addition, because the analyses concern family homelessness, the homeless families include only those that also report that the child lives with them all or most of the time.
Overall, the number of homeless respondents is small in these data. At the 1-year interview, 128 mothers living with children are identified as homeless. At the 3-year follow-up, 97 mothers living with children are identified as homeless. Among fathers who live with children but do not live with the child’s mother, very few are homeless (12 at the 1- year interview and 16 at the 3-year interview), so these families are not included in these analyses. They do include homeless families in which the mother and father live together. An indicator variable identifies these types of families in the analyses.
The current study also analyzes doubling up with friends or relatives
492 Social Service Review
as a form of homelessness. Doubling up is defined as living with family or friends or living in a house owned by family but, in either case, not paying rent. Among mothers living with children, 343 are coded as doubled up at the 1-year interview, and 223 are thusly coded at the 3- year interview. Among fathers living with children but not with the mother, few are doubled up (46 at the 1-year interview and 41 at the 3-year interview), so the fathers are not included in these analyses. Anal- yses do include doubled-up families composed of both the mother and the father.
In particular, the analyses focus on four homeless subsamples: mothers who are homeless ( ) or doubled up ( ) at the 1-yearn p 128 n p 343 interview, as well as those who are homeless ( ) or doubled upn p 97 ( ) at the 3-year interview. There is very little overlap across then p 223 interview waves; only 16 mothers are homeless and 88 mothers are dou- bled up at both the 1-year and the 3-year follow-up interviews. There are two comparison groups: all mothers in households with incomes at or below 50 percent of the federal poverty threshold at the 1-year in- terview and mothers living at or below 50 percent of the threshold at the 3-year interview.3 The assumption is that these groups of mothers, although not homeless or doubled up during the relevant interview period, are at risk of homelessness.
Individual-level measures used in these analyses come from the FFCW data. Demographic characteristics include the mother’s age at the sam- pled child’s birth, race and ethnicity (divided into four categories: white, black, Hispanic, and other), immigrant status, education (represented by whether the mother is a high school dropout or not), and marital status at the child’s birth.
Because most individual characteristics can change over time, moth- ers’ reports at the interview (or interviews) before the spell (with two exceptions) are used in these analyses. In addition to the mother’s marital status at the child’s birth, measures selected to capture family structure include the age of children, residential status of fathers, and household size. The age of the mother’s youngest child at the time of the interview is included to capture whether the mother has been preg- nant since the birth of the sampled child because pregnancy may affect such factors as the willingness of her family to provide her with a place to live. The age of the youngest child is measured at the same interview as the mother’s homeless and doubled-up status, but the other measures described below are derived from an interview (or interviews) before the homeless or doubled-up spell. The measure would not capture a pregnancy since the last interview if it only included information from the last interview.
Whether the mother and father live together at the time of the in- terview is included to capture whether the homeless spell involved the mother and children only or the mother, father, and children. For rea-
Urban Family Homelessness 493
sons similar to those described above, this variable is also measured at the same interview as the mother’s homeless or doubled-up status. The number of children at the time of the interview represents the house- hold size. The number includes the sampled child even at the baseline interview, right after the child’s birth.
Important influences on a mother’s material resources are her edu- cation and her employment status. Employment status at the baseline interview is derived from whether the mother worked at any time during pregnancy. The measure considers work during pregnancy because the baseline interview occurs at the time of the birth, when mothers cannot work temporarily. At the 1-year interview, the mother’s employment status reflects her working status at the time of the interview.
In addition, several other measures were selected to represent her material resources because these items capture with more detail a mother’s ability to maintain employment and manage her household’s budget at the time of interview; these include self-reported health status, depression, exposure to domestic violence, and presence of a drug prob- lem. Self-reported health status is measured by the mother’s response to a question about how she rates her general health at the time of the interview. Respondents choose one of five possible ratings: excellent, very good, good, fair, and poor. The scale takes the value one if health is reported to be excellent and five if it is reported to be poor. Depression is represented by a computed probability that a mother would be pos- itively diagnosed with depression.4 The probability is assessed using ques- tions about the presence and frequency of certain feelings and other symptoms, such as weight loss and trouble sleeping, over the 12 months prior to the interview. The list of feelings and symptoms was derived from the Composite International Diagnostic Interview—Short Form (Kessler et al. 1998). The mother’s exposure to domestic violence is assessed by her response to a question about whether she has been cut, bruised, or seriously hurt in a fight by the father of her child. This question is posed at the 1-year interview, but a follow-up question de- termines whether the violence occurred before the birth or after the birth. Finally, a drug problem is assumed if the mother responds that drinking or drugs have interfered with her work on a job or with her personal relationships in the 12 months prior to the interview.
Analyses also include measures of formal and informal social support. Informal social support is measured by the mother’s reported ability to get help from family, how long she has lived in her current (at the time of the interview) neighborhood, and how many times she has moved since the last interview. Three items gauge a family’s ability and will- ingness to support the respondent if she needed help in the year fol- lowing the interview: whether she could count on someone in her family to loan her $200, to provide a place to live, and to help with babysitting or child care. At the baseline interview, the mother is asked how long
494 Social Service Review
she has lived in her current neighborhood. The response provides an indicator of the extent of the network of friends she may have developed while in this neighborhood. From this survey item, an indicator is cre- ated to show whether the mother has lived in the neighborhood for 5 years or more. At the follow-up interviews, the number of residential moves is reported instead. In this case, a smaller number of moves would suggest a larger network of support. Formal support includes living in public housing, receipt of housing subsidies (e.g., Sec. 8 housing vouch- ers), and receipt of cash welfare (e.g., Temporary Assistance for Needy Families). The survey asks about receipt of cash welfare over the 12 prior months but asks about public housing residence and receipt of housing subsidies at the time of the interview.
To supplement these individual measures from the FFCW data, de- tailed city-level data are collected from various sources. These data re- flect the local economic environment, climate, housing affordability, housing availability, access to shelter beds, and antihomeless laws. In these analyses, the local economic environment is measured in 2000 for the analyses of the 1-year subsamples and in 2002 for the 3-year subsamples. The environment is measured by the unemployment rate for the city (U.S. Department of Labor, n.d.) and by the poverty rate for the county (or city, if available; U.S. Census Bureau, n.d.). The average maximum temperature in July, the average minimum temper- ature in January, and the average rainfall in inches, compiled by the U.S. National Oceanic and Atmospheric Administration (1992) using data from 1961 to 1990, represent the important features of a city’s climate in these analyses.5
Housing affordability and availability are measured by three statis- tics: (1) the city’s fair market rent for a two-bedroom apartment in 2005, (2) the percentage of apartments of all sizes in the city with rents at or below 30 percent of the city’s median family income in 2000, and (3) the city’s rental vacancy rate in 2000. Data on fair market rents are provided by HUD (2005). The rental vacancy rate is the per- centage of apartments that are vacant for the duration of the year in question. Both the percentage of apartments at or below 30 percent of the median family income and the rental vacancy rates are computed from data made publicly available by HUD (2000).
The number of shelter beds per 1,000 people in a city and the per- centage of those shelter beds that are reserved for families in 2004 were collected by calling local offices associated with HUD’s Continuum of Care initiative.6 These measures are designed to capture the extent to which cities have adopted policies designed to affirmatively support the homeless (particularly families). Finally, the number of antihomeless laws in 2005 and 2006 was compiled by the National Coalition for the Homeless, a nonprofit advocacy organization (Michael Stoops, executive director, National Coalition for the Homeless, personal communication,
Urban Family Homelessness 495
October 7, 2005; National Coalition for the Homeless and National Law Center for Homelessness and Poverty 2006). Examples of antihomeless laws include those that prohibit vagrancy, loitering, sitting, lying, camp- ing, sleeping, begging, urinating, or defecating in public places or bath- ing in particular public waters. The measure of antihomeless laws is designed to capture the extent to which cities have adopted policies designed to curb homelessness or stigmatize the homeless (including families). Although these antihomeless laws may be enacted to target single, homeless adults, the criminalization of homeless adults may spill over to the entire population and change the behavior of those who are at risk of becoming homeless.
Characteristics of the Homeless
Table 1 presents characteristics of mothers and their families for each of the four subsamples and the two comparison groups. The columns for the 1-year interview present characteristics reported at the baseline interview. The columns for the 3-year interview present characteristics reported at both the baseline and the 1-year interviews. For each char- acteristic, the table reports whether the means for the homeless and doubled-up samples differ to a statistically significant degree from the mean of the comparison group (i.e., those reporting income at 50 per- cent of poverty). Analyses also test whether the samples differ across interview waves (indicated by table note a).
Homeless mothers are statistically significantly less likely than the comparison groups (50 percent of poverty) to be immigrants and to be living with the child’s father at the time of the interview. Although differences between the homeless sample and the comparison group are often only statistically significant at one interview wave, homeless mothers are found to be more likely to have a drug problem, to have fair or poor health, and to have been physically hurt by their child’s father than are mothers in the comparison groups.7 They also have a higher probability of a depression diagnosis. Their relatives reportedly are less able to support them with loans, housing, or babysitting, and this is particularly true among the mothers who are homeless at the 1- year interview.8 Homelessness does not appear to be linked to race or ethnicity, marital status at the time of the child’s birth, age of the youn- gest child, the mother’s employment (during pregnancy or at the 1- year interview), receipt of housing subsidy, or receipt of cash welfare.
In general, these statistics suggest that the 1-year and 3-year subsam- ples of homeless families are very similar on such individual background characteristics as race, immigrant status, and educational attainment. A slightly smaller share of those homeless at the 3-year interview (23 per- cent) lives with the father compared to those homeless at the 1-year interview (30 percent). Compared with the 1-year subsample of home-
Table 1
Individual-Level Characteristics
At 1-Year Interview At 3-Year Interview
50% of Poverty
Doubled Up Homeless
50% of Poverty
Doubled Up Homeless
No. of observations 868 343 128 760 223 97 Mother’s age at child’s
birth (years) 23.7 21.7** 24.8� 23.7 22.9�,a 23.7 Mother is black 61 44** 61 64 39** 61 Mother is Hispanic 30 33 25 26a 34* 27 Mother is another race 3 4 3 2 5* 5 Mother is immigrant 17 13� 7** 13a 13 6** Mother is high school
dropout 57 40 50 58 33**,a 48� Mother was unmarried at
child’s birth 91 92 92 92 86**,a 92 Age of youngest child at 1-
year interview (months) 14.1 13.9 14.6
Living with father at 1-year interview 46 31** 30**
Age of youngest child at 3- year interview (months) 25.9a 29.3** 26.1
Living with father at 3-year interview 36 30 23**
Baseline interview: Children (no.) 2.5 1.5** 2.5 2.6 1.7**,a 2.3� Mother worked while
pregnant 64 72** 66 63 71* 59 Mother has a drug
problem 10 9 20** 11 11 14 Mother’s self-reported
health is fair or poor 11 6** 19* 12 6** 13 Mother has been hurt by
father 6 6 18** 6 4 11 Mother lives in public
housing 21 3** 19 22 4** 17 Mother receives housing
subsidy 13 4** 15 13 4** 13 Mother received cash
welfare 62 31** 56 57a 37** 56 Mother’s family would
loan $200 84 94** 73** 84 89*,a 82a Mother’s family would
house 88 96** 81� 87 96** 87 Mother’s family would
babysit 89 95** 87 89 95** 87 Mother has lived in
neighborhood ≥ 5 years 28 46** 23 27 46** 23
1-year interview: Children (no.) 2.6 1.7** 2.3* Mother currently
working 31 50** 33 Mother has a drug
problem 3 3 4 Mother’s self-reported
health is fair or poor 20 17 26
Urban Family Homelessness 497
Table 1 (Continued )
At 1-Year Interview At 3-Year Interview
50% of Poverty
Doubled Up Homeless
50% of Poverty
Doubled Up Homeless
Mother’s probability of depression diagnosis 16 14 26*
Mother has been hurt by father 5 8 11
Mother lives in public housing 29 7** 17**
Mother receives housing subsidy 13 4** 9
Mother received cash welfare 50 26** 56
Mother’s family would loan $200 74 80� 65�
Mother’s family would house 75 91** 71
Mother’s family would babysit 82 93** 74
Moves made by mother (no.) .8 .6* 1.1**
Note.—Unless otherwise specified, results are presented in percentages. Comparison group includes mothers with incomes at or below 50 percent of the federal poverty threshold.
a Comparison group composed of mothers in the corresponding group (50 percent of poverty, homeless, or doubled up) in the 1-year sample (not tested for the 1-year interview variables): .p ≤ .10
� .p ≤ .10 * .p ≤ .05 ** .p ≤ .01
less mothers, the 3-year subsample of homeless mothers also reports lower rates of drug problems, fair or poor health, and being physically hurt by the father of her child at the baseline interview. The 3-year subsample reports greater access to loans and housing from family (two of the three measured components of family support). Still, differences across the 1-year and 3-year homeless subsamples are statistically sig- nificant in only one case and appear to be limited in size and scope in general.
In contrast, the table suggests that mothers who doubled up differ in many ways from homeless mothers and from the comparison group of mothers. In addition, doubled-up mothers differ to some degree across interview waves, albeit, not in direction. Compared with mothers who are homeless or in the comparison group, doubled-up mothers are slightly younger, are less likely to be black, and have fewer children. Doubled-up mothers are more likely to be Hispanic or of another race and less likely to be a high school dropout. Mothers in the doubled-up subsamples are more likely to be working and less likely to report having fair or poor health. Doubling up does not appear to be linked to reports
498 Social Service Review
Table 2
City-Level Characteristics
Mean Min Max
Unemployment rate in 2000 5 3 8 Unemployment rate in 2002 7 4 13 Poverty rate in 2000 14 7 19 Poverty rate in 2002 15 8 21 Average maximum temperature in July (degrees Fahrenheit) 86 70 95 Average minimum temperature in January (degrees Fahrenheit) 27 12 45 Average rainfall (inches) 38 14 51 Fair market rent (dollars) 876 571 1,342 Apartments with median family incomerents ! 30% 19 6 34 Rental vacancy rate 6 2 11 Shelter beds per 1,000 (no.) 2 .9 6.9 Shelter beds for families 41 10 64 Antihomeless laws (no.) 6 3 9
Note.—Unless otherwise specified, results are presented in percentages. The 20 Fragile Families and Child Wellbeing study cities are Austin, TX; Baltimore, MD; Boston, MA; Chicago, IL; Corpus Christi, TX; Detroit, MI; Indianapolis, IN; Jacksonville, FL; Milwaukee, WI; Nashville, TN; New York, NY; Newark, NJ; Norfolk, VA; Oakland, CA; Philadelphia, PA; Pittsburgh, PA; Richmond, VA; San Antonio, TX; San Jose, CA; and Toledo, OH.
of drug problems, probability of depression diagnosis, or experience of domestic violence. Fewer doubled-up mothers report living in public housing, receiving a housing subsidy, and receiving welfare than do mothers who are homeless or in the comparison group. Mothers in the doubled-up subsamples are more likely to have relatives willing and able to lend money, house, and babysit. Compared with those in the homeless and comparison groups, doubled-up mothers are also more likely to have lived in their neighborhood 5 years or more.
Table 2 displays city-level characteristics that may influence the prob- ability of family homelessness. The 20 FFCW cities vary greatly in these characteristics. The economic strength of the cities ranges from Corpus Christi, TX, which has the highest poverty rate in both 2000 and 2002 (19 and 21 percent, respectively), to San Jose, CA, which has the lowest poverty rate in both years (7 and 8 percent, respectively). The climate may have a large influence on the homeless population. The data there- fore include three cities in Texas (Austin, Corpus Christi, and San An- tonio) that have very high year-round temperatures, as well as cities like Milwaukee, WI, which has an average minimum temperature in January of 12 degrees Fahrenheit.
Table 2 suggests that there is substantial variation in all three variables that measure housing affordability and availability: fair market rent, the percentage of apartments for which rent is less than 30 percent of the area’s median family income, and the rental vacancy rate.9 The table also reports summary statistics for the three measures of city-level home- less policy: the number of shelter beds per 1,000 people in the city, the
Urban Family Homelessness 499
percent of shelter beds that are reserved for families, and the number of antihomeless laws that a city has enacted.
Predictors of Homelessness
The effect of all of these individual- and city-level variables on the prob- ability of being homeless or doubled up is estimated in analyses reported in tables 3 and 4. Multinomial logit estimation is used. The first columns report the estimated effect of each independent variable on doubling up versus being in the comparison group or the homeless group at the 1-year interview. The second set of columns report the estimated effect of the variables on having a homeless spell versus being in the com- parison group or the doubled-up group at the 1-year interview. The coefficients in both sets of columns are from a single regression.
In table 3, the dependent variable has three values that indicate whether an observation is in the comparison group, the doubled-up group, or the homeless group at 1 year, and the independent variables include the characteristics of the mother at baseline as well as city var- iables from 2000 (if possible). The R 2 statistic suggests that about 22 percent of the variation in homelessness and doubling up is explained with these explanatory variables. Otherwise, results suggest that the like- lihood of homelessness increases with the age of the mother. Immigrant status is estimated to reduce a mother’s risk for homelessness. Living with the father and the number of children are negatively associated with the mother’s risk of being homeless. Having been physically hurt by the child’s father is positively associated with, and self-reported health status is negatively related to, increases in a mother’s risk of homeless- ness. Family support (represented by having relatives who would lend money, house, and babysit) and living in a neighborhood for 5 years or more are negatively associated with the risk of homelessness. Even if analyses control for these individual characteristics, several city-level characteristics have statistically significant effects. In particular, the prob- ability of homelessness at the 1-year interview is associated with increases in the fair market rent, the scarcity of affordable housing units (i.e., apartments with rents less than 30 percent of median family income), and the rate of rental vacancies, although the effects are only statistically significant at the 10 percent level for two of the three effects.
Black and Hispanic mothers are estimated to be statistically signifi- cantly less likely to be doubled up than are white mothers or mothers of another race. Living with the father and the number of children are negatively associated with the mother’s risk of doubling up. Family sup- port (represented by having relatives who would lend money, house, and babysit) and living in a neighborhood 5 years or more are positively associated with doubling up. Receipt of public housing, housing sub- sidies, and cash welfare are negatively related to the probability of dou-
500 Social Service Review
Table 3
Predicting Homelessness and Doubling Up at 1 Year after Birth
Doubled Up Homeless
Coefficient SE Coefficient SE
Mother’s age at child’s birth .010 .016 .077** .018 Mother is black �.733** .269 �.421 .366 Mother is Hispanic �.554* .226 �.329 .592 Mother is another race .167 .599 .307 .741 Mother is immigrant �.557 .356 �1.829** .567 Mother is high school dropout �.267 .187 �.064 .224 Mother was unmarried at child’s birth �.093 .327 �.457 .401 Age of youngest child at 1-year interview �.016 .024 .036 .039 Living with father at 1-year interview �.622* .252 �.722** .256 No. of children at baseline �.717** .102 �.419** .146 Mother worked while pregnant �.154 .124 .005 .196 Mother has a drug problem at baseline �.089 .287 .245 .328 Mother’s self-reported health status at baseline
( , )1 p excellent 5 p poor �.083 .081 .280** .087 Mother has been hurt by child’s father before
birth .145 .487 1.064** .319 Mother lives in public housing at baseline �2.075** .320 �.041 .237 Mother receives housing subsidy at baseline �.672� .369 .052 .326 Mother receives cash welfare at baseline �.907** .170 �.373 .267 Mother’s family would help at baseline (3 p loan
$200, house, and babysit) .309** .090 �.282** .108 Mother has lived in years atneighborhood ≥ 5
baseline .586** .192 �.505* .249 Unemployment rate in 2000 �.034 .112 .010 .081 Poverty rate in 2000 .017 .025 �.001 .031 Average maximum temperature in July .021 .020 �.018 .018 Average minimum temperature in January .030 .019 �.016 .016 Average rainfall .005 .015 .013 .017 Fair market rent (log) �.013 1.071 1.812� 1.035 % apartments with median familyrents ! 30%
income .007 .022 �.064* .027 Rental vacancy rate �.016 .063 .123� .075 No. of shelter beds per 1,000 �.020 .132 .114 .090 % shelter beds for families �.009 .012 �.002 .010 No. of antihomeless laws �.029 .063 .092 .070 Observations 1,262 R 2 .2219
Note.—Comparison group is households at 50% of the poverty line at the 1-year interview. Multinomial logits are used. Coefficients indicate the estimated effect of the independent variable on being doubled up or homeless over neither. Robust standard errors reported.
� .p ≤ .10 * .p ≤ .05 ** .p ≤ .01
bling up. If analyses control for these individual characteristics, no city- level characteristics have statistically significant effects on doubling up.
Table 4 reports the results from the regression concerning the 3-year interview sample. In these analyses, the independent variables are char- acteristics of the mother from the 1-year interview and city characteristics from 2002 (if possible). The R 2 indicates that 21 percent of the variation in the dependent variables is explained. As above, results suggest that
Urban Family Homelessness 501
Table 4
Predicting Homelessness and Doubling Up at 3 Years after Birth
Doubled Up Homeless
Coefficient SE Coefficient SE
Mother’s age at child’s birth .003 .024 �.019 .030 Mother is black �.690* .299 .573 .407 Mother is Hispanic �.223 .381 .744 .580 Mother is another race .508 .589 1.527** .487 Mother is immigrant �.274 .337 �1.382* .550 Mother is high school dropout �.607* .304 �.545* .250 Mother was unmarried at child’s birth �.450� .269 �.271 .364 Age of youngest child at 3-year interview .017* .008 �.003 .012 Living with father at 3-year interview �.401� .226 �.576* .291 No. of children at 1-year interview �.502** .125 �.112 .158 Mother currently working at 1-year interview .283 .225 .010 .227 Mother has a drug problem at 1-year interview .370 .564 �.165 .641 Mother’s self-reported health status at 1-year in-
terview ( , )1 p excellent 5 p poor �.137* .064 .115 .086 Mother’s probability of depression diagnosis at 1-
year interview �.044 .203 .475� .282 Mother has been hurt by child’s father between
birth and 1-year interview .432 .285 .511 .525 Mother lives in public housing at 1-year interview �1.504** .326 �.858* .373 Mother receives housing subsidy at 1-year
interview �.905� .474 �.909� .492 Mother receives cash welfare at 1-year interview �.488** .174 .129 .186 Mother’s family would help at 1-year interview
( $200, house, and babysit)3 p loan .169* .076 �.202 .135 No. of moves made by mother between birth and
1-year interview �.233� .138 .259* .102 Unemployment rate in 2002 �.125* .057 .007 .106 Poverty rate in 2002 �.041 .034 �.004 .049 Average maximum temperature in July .075** .014 �.023 .039 Average minimum temperature in January .034** .013 �.009 .037 Average rainfall .007 .023 .019 .031 Fair market rent (log) �1.049 .885 2.517 2.482 % apartments with median familyrents ! 30%
income .045* .021 �.076� .039 Rental vacancy rate �.172* .073 .183 .148 No. of shelter beds per 1,000 .079 .084 .274 .217 % shelter beds for families .021** .005 .005 .024 No. of antihomeless laws �.268** .052 .110 .155 Observations 1,001 R 2 .2113
Note.—Comparison group is households at 50% of the poverty line at the 3-year interview. Multinomial logits are used. Coefficients indicate the estimated effect of the independent variable on being doubled up or homeless over neither. Robust standard errors reported.
� .p ≤ .10 * .p ≤ .05 ** .p ≤ .01
immigrant status and living with the father are negatively associated with homelessness. Moving residences frequently is positively related to homelessness. In addition, results suggest that mothers in the other race category are more likely to be homeless than are mothers who are white, black, or Hispanic. Being a high school dropout is negatively associated with the likelihood of being homeless in this low-income population.
502 Social Service Review
The mother’s risk of homelessness rises with the probability that she will have a depression diagnosis. Results suggest that receipt of public housing or a housing subsidy at the 1-year interview is negatively and statistically significantly associated with the probability of being homeless at the 3-year interview. Finally, homelessness is inversely related to the percentage of apartments with rents lower than 30 percent of the me- dian family income.
Results suggest that black mothers are statistically significantly less likely to be doubled up than mothers who are white, Hispanic, or of another race. Being a high school dropout and being unmarried at the cohort member’s birth are negatively associated with the probability of doubling up at the 3-year interview. So too, the likelihood of doubling up increases with the age of the youngest child and is negatively related to both living with the father and the number of children in the house- hold. Poor self-reported health is negatively associated with the prob- ability that mothers will double up. Having relatives who would lend money, house, or babysit is estimated to be positively associated with the probability of doubling up. It is negatively associated with the num- ber of moves made between the child’s birth and the 1-year interview. Receipt of public housing, housing subsidies, and cash welfare are neg- atively related to doubling up.
The city variables have a higher estimated effect on doubling up at the 3-year interview than at the 1-year interview. In particular, the probability of doubling up is estimated to be negatively associated with unemployment rates, a city’s rental vacancy rate, and the number of antihomeless laws. It is positively associated with the percentage of af- fordable (i.e., apartments with rents of less than 30 percent of the city’s median family income) housing units and with the percentage of shelter beds reserved for families.
The reason for the differences in results across interview waves is not clear. One of the main differences between the samples is that the 1- year interview is closer in time to the dramatic event of a birth. Thus, analyses control for the age of the youngest child in the family. Results do suggest that the age of the youngest child at the 3-year interview is associated with doubling up but not with homelessness. Further analyses examine the interactions between the age of the youngest child and the other independent variables that are found to be statistically sig- nificant in table 4. However, the interactions are not statistically signif- icant (and thus not included in the specification shown).
Another difference across interview waves is the state of the economy. The U.S. economy went into recession in 2001. Table 2 shows that unemployment and poverty rates were higher in 2002 than in 2000. However, control variables account for economic differences between the samples. Results reported in table 1 indicate that, despite the eco- nomic conditions, there were fewer homeless and doubled-up families
Urban Family Homelessness 503
in the sample at the 3-year interview (9 and 21 percent, respectively) than there were at the 1-year interview (10 and 26 percent).
At least two analytic issues are of concern. First, some of the inde- pendent variables may not predict homelessness because they are mea- sured with more error than others. In particular, receipt of welfare and employment vary dramatically over time in this low-income population, and one snapshot may miss important features of the population’s ex- perience with welfare and employment in predicting their risk of home- lessness. Because the questions are posed at a point in time, stable variables, like a person’s race and immigrant status, will be reliable proxies of the true value, and variables that can change over time, like a person’s welfare status and employment status, may measure these characteristics with error, making them less likely to be statistically sig- nificantly different from zero. To test whether this might be the case, more stable measures of welfare status and working status are created from two waves of data instead of one. These 2-period measures have the same estimated effects on homelessness as the 1-period measures. In addition, welfare status measured in only 1 period has a statistically significantly negative relation to the probability of doubling up in both the 1-year and 3-year samples. This finding indicates that these variables have predictive power. Still, it may be the case that these unstable traits suffer from attenuation bias and should therefore be interpreted with care.
Second, the homeless policy variables may act as an indicator for cities with high levels of homelessness. That is, the likelihood of funding homeless shelters and passing antihomeless laws may rise with the size of the homeless population. Thus, in analyses that control for effects of homeless policies, these variables could absorb all of the effects ac- tually attributable to other factors. Because of this, the regressions shown in tables 3 and 4 were reestimated to omit potentially problematic var- iables. Results (not shown) suggest that if the homeless policies are omitted, the likelihood that families experience a homeless spell rises with the fair market rent. No other effects rise to the level of statistical significance. Thus, it appears that, for the most part, the inclusion of the number of shelter beds, the percentage of beds reserved for families, and the number of antihomeless laws does not distort the estimated effects of the other explanatory factors.
Discussion and Conclusion
The analytic approach adopted here is to measure some of the many individual and structural factors that may be associated with the in- creased risk of becoming homeless or doubled up while also measuring the characteristics thought to protect families from such hardships. The design of the FFCW provides the necessary data, enabling researchers
504 Social Service Review
to overcome the limitations of studies that focus only on individual or structural factors. It also avoids the limitations of research that lacks an adequate comparison group in considering families with experience in homelessness.
The current approach establishes the relative importance of factors frequently thought to shape spells of homelessness and doubling up among low-income families. In so doing, it provides the kind of infor- mation that can help untangle the complex matrix of risk factors. This type of analysis can also help policy makers to prioritize strategies that promise the greatest reductions in the number of families living in shelters or on the street.
The analysis highlights the primacy of individual factors over struc- tural constraints in predicting homelessness. In particular, poor general and mental health, domestic violence, and residential mobility are found to predict the likelihood of homelessness even if analyses control for city-level variation in housing affordability, local economic conditions, climate, shelter availability, and antihomeless laws. If structural factors are held constant, the probability of homelessness is negatively related to immigrant status, living in public housing, and housing subsidy re- ceipt, as well as to having relatives who would support the mother by lending money, housing, and babysitting. Local housing conditions seem to be associated with the risk of being homeless or doubled up, and these relations seem to be independent of individual- and household- level sociodemographic characteristics, but the statistical significance of the relation is only marginal. In order to place these findings within the context of the homelessness literature, the results of the current analysis are examined below with respect to the three dominant factors thought to precipitate homelessness: the lack of material resources, fam- ily and social support, and local housing market conditions.
Findings are mixed concerning the role in homelessness of material resources. Results fail to suggest that the lack of material resources, proxied by employment status and low educational attainment, increases the risk of homelessness among low-income families. Similarly, analyses find no protective effects of cash welfare benefits. Moreover, local eco- nomic conditions, such as the local poverty and unemployment rates, provide little leverage in explaining why low-income families become homeless. However, the selection of comparison groups (in this study, families that are not homeless or doubled up but that are at risk of homelessness) is based on household income, such that the comparison group is poorer and less educated, on average, than the homeless or doubled-up families are. In addition, the FFCW sample is dispropor- tionately made up of unmarried parents, who are more likely to reside in low- and moderate-income households. This sample selection limits the variation in measures of the sample population’s economic resources (e.g., income). Compared with a representative sample of all U.S. house-
Urban Family Homelessness 505
holds, this select sample also likely has a smaller observed effect of economic resources on measures of material hardship, like homeless- ness. As a result, it is not entirely surprising that analyses fail to uncover statistically significant relations between economic resources and home- lessness.
However, physical health, mental health, and domestic violence (likely indicators of material hardship) are found to play roles in homelessness. Poor physical or mental health may trigger homelessness through its effect on work and, thus, on material hardship. Interestingly, relevant relations are observed soon after the birth of the sampled child (except for mental health, which is not measured at baseline) but disappear over time. Perhaps the birth of a new child leaves poor families highly vulnerable to crises. Alternately, a child’s birth may prompt the mother to reconsider a relationship marked by domestic violence, and this re- consideration may result in an increased risk of homelessness. The rate of self-reported domestic abuse among homeless respondents decreases over the sample periods (from 18 to 11 percent).
In addition to domestic violence and poor health, household structure appears to help explain family homelessness. Although marriage does not seem to provide protective effects against family homelessness, the odds that a family will become homeless are negatively related to living with the sample child’s father. It seems plausible that live-in fathers limit the potentially devastating consequences of severe economic hardship, whether that hardship results from the loss of a job due to economic conditions, from poor health, or from some other reason.
The potential protective effects that fathers provide against home- lessness can be viewed as a form of informal social support. Findings provide additional evidence that informal social support explains why some poor families become homeless and others do not, particularly in the first year after a birth. Access to relatives who can provide a small loan, child care, or a place to live is found to be negatively and statis- tically significantly associated with the odds of becoming homeless at the 1-year interview. Understanding informal social support is a central factor in understanding the family homelessness problem, and the im- portance of such support spills over time into formal (or institutional) types of assistance. In particular, housing subsidies are found to provide protection against homelessness at the 3-year interview. It would appear that poor families faced with a personal or economic crisis are able to avoid homelessness in part because of low-income housing assistance.
Local housing market conditions represent the final dominant factor found in the literature and are thought to influence homelessness. The current study’s findings suggest that housing costs and lack of affordable units are positively associated with the odds of being homeless; however, the estimated effect of high rent in a given city is only marginally sta- tistically significant, and the coefficient linked to affordability is relatively
506 Social Service Review
small. To the extent that local housing market conditions are related to the risk of family homelessness, the conditions appear to be less important than household composition or social support.
These analyses have several limitations. First, the measure of home- lessness is somewhat crude and does not capture variation in the severity of homelessness spells. Sleeping in a homeless shelter for a few nights is substantially different from sleeping on the street for months; however, the FFCW is unable to distinguish among levels of severity in home- lessness spells or to measure the number of homelessness spells. The relatively broad definition of homelessness used in this analysis likely creates a more heterogeneous sample of homeless families than if, say, only families who experience multiple, long-term spells of homelessness were examined. As a result, it may not be possible to isolate particular factors associated with chronic homelessness. Second, the number of homeless families in the subsamples limits the precision of the estimates. The robustness of the findings would likely improve with a larger sample of homeless families. Third, this study is restricted to the 20 cities in the FFCW. It is unclear whether the findings can be generalized beyond these places. This threat to external validity may be overstated given the mix of cities by region, size, and level of deprivation; however, it is important that these results not be interpreted beyond the study sample. Fourth, the assumed causal direction between the dependent and in- dependent variables may be reversed. An attempt is made to limit reverse causality problems by measuring health, for example, prior to the re- ported homeless spell, but there may have been homeless spells prior to the health assessment. Thus, the findings are interpreted to suggest that poor health increases the risk of homelessness, but it is also likely that chronic homelessness affects physical health. This problem can be addressed by using future waves of data from the FFCW.
Although future research will have to overcome the limitations, the analytic approach adopted here is an innovative strategy for the study of family homelessness (and homelessness in general). It provides a framework for improving knowledge about the problem, as well as a coherent and comprehensive approach to the study of it. The approach may also give policy makers the type of knowledge and understanding they need to craft effective interventions designed to keep at-risk families from becoming homeless.
These results suggest that policy makers need to account for the multiplicity of factors that lead to family homelessness at individual and community levels. Specifically, results suggest that family homelessness services should consider targeting prevention efforts on native-born mothers who do not live with the father of their children, those who may have health or safety concerns, and those who do not have mean- ingful family or institutional social support. In addition, municipal gov- ernments should not be overly concerned that the availability of shelter
Urban Family Homelessness 507
space will increase homelessness among families. So too, advocates and policy makers should be mindful that local labor and housing market conditions are not strongly linked to family homelessness.
References Bassuk, Ellen L., John C. Buckner, Jennifer N. Perloff, and Shari S. Bassuk. 1998. “Prev-
alence of Mental Health and Substance Use Disorders among Homeless and Low- Income Housed Mothers.” American Journal of Psychiatry 155 (11): 1561–64.
Bassuk, Ellen L., John C. Buckner, Linda F. Weinreb, Angela Browne, Shari S. Bassuk, Ree Dawson, and Jennifer N. Perloff. 1997. “Homelessness in Female-Headed Families: Childhood and Adult Risk and Protective Factors.” American Journal of Public Health 87 (2): 241–48.
Bassuk, Ellen L., and Stephanie Geller. 2006. “The Role of Housing and Services in Ending Family Homelessness.” Housing Policy Debate 17 (4): 781–806.
Bassuk, Ellen L., and Lynn Rosenberg. 1988. “Why Does Family Homelessness Occur? A Case-Control Study.” American Journal of Public Health 78 (7): 783–88.
Bassuk, Ellen L., Linda F. Weinreb, John C. Buckner, Angela Browne, Amy Salomon, and Shari S. Bassuk. 1996. “The Characteristics and Needs of Homeless and Low-Income Mothers.” Journal of the American Medical Association 276 (8): 640–46.
Benda, Brent B. 1990. “Crime, Drug Abuse and Mental Illness: A Comparison of Homeless Men and Women.” Journal of Social Service Research 13 (3): 39–60.
Bloom, Dan, Mary Farrell, and Barbara Fink, with Diana Adams-Ciardullo. 2002. “Welfare Time Limits: State Policies, Implementation, and Effects on Families.” Report to the U.S. Department of Health and Human Services, Administration for Children and Families, Office of Planning, Research, and Evaluation (contract no. 282-00-0014, task no. 002). Manpower Demonstration Research Corporation, New York.
Bohanon, Cecil. 1991. “The Economic Correlates of Homelessness in Sixty Cities.” Social Science Quarterly 72 (4): 817–25.
Burt, Martha R. 2001a. “Homeless Families, Singles, and Others: Findings from the 1996 National Survey of Homeless Assistance Providers and Clients.” Housing Policy Debate 12 (4): 737–80.
———. 2001b. “What Will It Take to End Homelessness?” Issue Brief, September. Urban Institute, Washington, DC.
Burt, Martha R., Laudan Y. Aron, Tobe Douglas, Jesse Valente, Edgar Lee, and Britta Iwen. 1999. “Homelessness: Programs and the People They Serve; Findings of the National Survey of Homeless Assistance Providers and Clients.” Technical report. U.S. De- partment of Housing and Urban Development, Washington, DC.
Culhane, Dennis P., Stephen Metraux, Jung Min Park, Maryanne Schretzman, and Jesse Valente. 2007. “Testing a Typology of Family Homelessness Based on Patterns of Public Shelter Utilization in Four U.S. Jurisdictions: Implications for Policy and Program Planning.” Housing Policy Debate 18 (1): 1–28.
Dworsky, Amy Lynn, and Irving Piliavin. 2000. “Homeless Spell Exits and Returns: Sub- stantive and Methodological Elaborations on Recent Studies.” Social Service Review 74 (2): 193–213.
Early, Dirk W. 1998. “The Role of Subsidized Housing in Reducing Homelessness: An Empirical Investigation Using Micro-Data.” Journal of Policy Analysis and Management 17 (4): 687–96.
———. 2004. “The Determinants of Homelessness and the Targeting of Housing Assis- tance.” Journal of Urban Economics 55 (1): 195–214.
Elliott, Marta, and Lauren J. Krivo. 1991. “Structural Determinants of Homelessness in the United States.” Social Problems 38 (1): 113–31.
Glomm, Gerhard, and Andrew John. 2002. “Homelessness and Labor Markets.” Regional Science and Urban Economics 32 (5): 591–606.
Goodman, Lisa A. 1991. “The Prevalence of Abuse among Homeless and Housed Poor Mothers: A Comparison Study.” American Journal of Orthopsychiatry 61 (4): 489–500.
HUD (U.S. Department of Housing and Urban Development). 2000. “Comprehensive Hous- ing Affordability Strategy (CHAS) Data.” http://socds.huduser.org/chas/index.html.
508 Social Service Review
———. 2005. “County Level Data File.” Fair Market Rent data. http://www.huduser.org/ datasets/fmr/fmr2005f/FMR_Area_2005_FMRs_CountyLevel.xls.
HUD (U.S. Department of Housing and Urban Development), Office of Community Planning and Development. 2007. The Annual Homeless Assessment Report to Congress. Report, February. Washington, DC: HUD, Office of Community Planning and De- velopment.
Imbroscio, David L. 2008. “‘United and Actuated by Some Common Impulse of Passion’: Challenging the Dispersal Consensus in American Housing Policy Research.” Journal of Urban Affairs 30 (2): 111–30.
Institute for Family and Social Responsibility. 2000. “Final Report on Client Study of the Community Social Services Study of the Impact of Indiana’s Welfare Reforms.” Report no. 00-05 (September 1). Indiana University, Institute for Family and Social Respon- sibility, Bloomington.
Jencks, Christopher. 1994. The Homeless. Cambridge, MA: Harvard University Press. Kessler, Ronald C., Gavin Andrews, Daniel K. Mroczek, Bedirhan T. Ustun, and Hans-
Ulrich Wittchen. 1998. “The World Health Organization Composite International Diagnostic Interview Short-Form (CIDI-SF).” International Journal of Methods in Psychi- atric Research 7 (4): 171–85.
Lee, Barrett A., Townsand Price-Spratlen, and James W. Kanan. 2003. “Determinants of Homelessness in Metropolitan Areas.” Journal of Urban Affairs 25 (3): 335–55.
Lehmann, Erika R., Philip H. Kass, Christiana M. Drake, and Sara B. Nichols. 2007. “Risk Factors for First-Time Homelessness in Low-Income Women.” American Journal of Or- thopsychiatry 77 (1): 20–28.
Letiecq, Bethany L., Elaine A. Anderson, and Sally A. Koblinsky. 1998. “Social Support of Homeless and Housed Mothers: A Comparison of Temporary and Permanent Housing Arrangements.” Family Relations 47 (4): 415–21.
Loprest, Pamela J. 1999. “How Families That Left Welfare Are Doing: A National Picture.” New Federalism: National Survey of America’s Families, report no. B-1. Urban Insti- tute, Washington, DC.
Ma, Christine T., Lauren Gee, and Margot B. Kushel. 2008. “Associations between Housing Instability and Food Insecurity with Health Care Access in Low-Income Children.” Ambulatory Pediatrics 8 (1): 50–57.
Main, Thomas J. 1998. “How to Think about Homelessness: Balancing Structural and Individual Causes.” Journal of Social Distress and the Homeless 7 (1): 41–54.
McPherson, Miller, Lynn Smith-Lovin, and Matthew E. Brashears. 2006. “Social Isolation in America: Changes in Core Discussion Networks over Two Decades.” American So- ciological Review 71 (3): 353–75.
National Coalition for the Homeless and the National Law Center on Homelessness and Poverty. 2006. “A Dream Denied: The Criminalization of Homelessness in U.S. Cities.” Report, January. http://www.nationalhomeless.org/publications/reports.html.
Phelan, Jo C., and Bruce G. Link. 1999. “Who Are ‘the Homeless’? Reconsidering the Stability and Composition of the Homeless Population.” American Journal of Public Health 89 (9): 1334–38.
Quigley, John M., Steven Raphael, and Eugene Smolensky. 2001. “Homeless in America, Homeless in California.” Review of Economics and Statistics 83 (1): 37–51.
Salomon, Amy, Shari S. Bassuk, and Margaret G. Brooks. 1996. “Patterns of Welfare Use among Poor and Homeless Women.” American Journal of Orthopsychiatry 66 (4): 510–25.
Shinn, Marybeth, Jamie Gottlieb, Jessica L. Wett, Ajay Bahl, Arnold Cohen, and Deborah Baron Ellis. 2007. “Predictors of Homelessness among Older Adults in New York City— Disability, Economic, Human, and Social Capital and Stressful Events.” Journal of Health Psychology 12 (5): 696–708.
Shinn, Marybeth, James R. Knickman, and Beth C. Weitzman. 1991. “Social Relationships and Vulnerability to Becoming Homeless among Poor Families.” American Psychologist 46 (1): 1180–87.
Shinn, Marybeth, Beth C. Weitzman, Daniela Stojanovic, James R. Knickman, Lucila Jiménez, Lisa Duchon, Susan James, and David H. Krantz. 1998. “Predictors of Home- lessness among Families in New York City: From Shelter Request to Housing Stability.” American Journal of Public Health 88 (11): 1651–57.
Shlay, Anne B., and Peter H. Rossi. 1992. “Social Science Research and Contemporary Studies of Homelessness.” Annual Review of Sociology 18:129–60.
Urban Family Homelessness 509
Sosin, Michael R. 1989. “Homelessness in Chicago: A Study Sheds New Light on an Old Problem.” Public Welfare 47 (1): 22–28.
Stainbrook, Kristin A., and John Hornik. 2006. “Similarities in the Characteristics and Needs of Women with Children in Homeless Family and Domestic Violence Shelters.” Families in Society 87 (1): 53–62.
Tischler, Victoria A., Alison Rademeyer, and Panos Vostanis. 2007. “Mothers Experiencing Homelessness: Mental Health, Support and Social Care Needs.” Health and Social Care in the Community 15 (3): 246–53.
Tischler, Victoria A., and Panos Vostanis. 2007. “Homeless Mothers: Is There a Relationship between Coping Strategies, Mental Health and Goal Achievement?” Journal of Com- munity and Applied Social Psychology 17 (2): 85–102.
Toohey, Siobhan M., Marybeth Shinn, and Beth C. Weitzman. 2004. “Social Networks and Homelessness among Women Heads of Household.” American Journal of Community Psychology 33 (1–2): 7–20.
Urban Institute. 2004. “How Are Families from America’s Worst Public Housing Faring?” Transcript of First Tuesdays forum, October 5. Urban Institute, Washington, DC. http:// www.urban.org/publications/900750.html.
U.S. Census Bureau. n.d. “Small Area Poverty and Income Estimates.” County and city- level poverty rate data. http://www.census.gov/hhes/www/saipe/saipe.html (accessed August 5, 2008).
U.S. Department of Labor. n.d. Local Area Unemployment Statistics database. http:// data.bls.gov/PDQ/outside.jsp?surveypla (accessed August 5, 2008).
U.S. National Oceanic and Atmospheric Administration, National Climatic Data Center. 1992. “Climatography of the United States, Number 81.” http://www.ncdc.noaa.gov/ oa/climate/normals/usnormalshist.html.
Venkatesh, Sudhir Alladi, Isil Çelimli, Douglas Miller, Alexandra Murphy, and Beauty Turner. 2004. “Chicago Public Housing Transformation: A Research Report.” Working paper. Columbia University, Center for Urban Research and Policy, New York.
Weitzman, Beth C. 1989. “Pregnancy and Childbirth: Risk Factors for Homelessness?” Family Planning Perspectives 21 (4): 175–78.
Wong, Yin-Ling Irene, Dennis P. Culhane, and Randall Kuhn. 1997. “Predictors of Exit and Reentry among Family Shelter Users in New York City.” Social Service Review 71 (3): 441–62.
Wong, Yin-Ling Irene, Irving Piliavin, and Bradley R. Entner Wright. 1998. “Residential Transitions among Homeless Families and Homeless Single Individuals: A Comparison Study.” Journal of Social Service Research 24 (1–2): 1–27.
Wood David L., Burciaga Valdez, Toshi Hayashi, and Albert Shen. 1990. “Health of Home- less Children and Housed, Poor Children.” Pediatrics 86 (6): 858–66.
Wright, Bradley R. Entner, Avshalom Caspi, Terrie E. Moffitt, and Phil A. Silva. 1998. “Factors Associated with Doubled-Up Housing—a Common Precursor to Homeless- ness.” Social Service Review 72 (1): 92–111.
Notes
The authors thank the Bendheim-Thoman Center for Research on Child Wellbeing at Princeton University, which is supported by grant 5 R01 HD36916 from the National Institute of Child Health and Human Development for restricted access to the Fragile Families and Child Wellbeing study.
1. The 20 cities include Austin, TX; Baltimore, MD; Boston, MA; Chicago, IL; Corpus Christi, TX; Detroit, MI; Indianapolis, IN; Jacksonville, FL; Milwaukee, WI; Nashville, TN; New York, NY; Newark, NJ; Norfolk, VA; Oakland, CA; Philadelphia, PA; Pittsburgh, PA; Richmond, VA; San Antonio, TX; San Jose, CA; and Toledo, OH.
2. The response rates vary slightly by marital status. For married mothers, 82 percent responded to the baseline interview, 91 percent of the baseline sample responded to the 1-year follow-up, and 89 percent responded to the 3-year follow-up. For unmarried mothers, 87 percent responded at baseline, 90 percent responded at 1 year, and 88 percent re- sponded at 3 years. For married fathers, 89 percent responded to the baseline interview, 82 percent of the baseline sample responded to the 1-year follow-up, and 82 percent
510 Social Service Review
responded to the 3-year follow-up. For unmarried fathers, 75 percent responded at base- line, 70 percent responded at 1 year, and 68 percent responded at 3 years. Because attrition rates are low, particularly for the mother interviews, the differences in the sample across the waves are negligible (detailed attrition analysis is available from the authors upon request).
3. The U.S. federal poverty threshold from the interview year is applied to the family’s income in that year to assign respondents to the comparison groups.
4. An assessment of mental health status is not taken at the baseline interview and thus is only available at the 1-year interview.
5. Climate averages using data from 1971 to 2000 were not free to the public at the time of this analysis. The differences in these overlapping 30-year averages are likely to be very small.
6. Those data are available upon request from the authors. More current data are available online at http://www.hud.gov/offices/cpd/homeless/local/index.cfm.
7. The percentage of mothers who report fair or poor health is shown in table 1, but the five-scale self-reported health status, whose mean is less meaningful, is used in the regression analysis.
8. All three questions about family support are reported in table 1. An index combining the three questions is used in the regression analysis to capture the degree of family support.
9. The logarithm of fair market rent is used in the regression analyses.
