Characteristics and Dynamics of Homeless Families with Children

A Reanalysis of the Fragile Families and Child Well-Being Study

[ Main Page of Report | Contents of Report ]

Contents

1. Introduction
2. Methodology
   1. Database Description
   2. Defining the Sample for Reanalysis
   3. Creating and Describing Residential Outcome Groups
   4. Potential Risk and Protective Factors
3. Descriptive Analyses and Results
4. Predicting Residential Stability and Homelessness
5. Discussion
   1. Summary of Results
   2. Caveats and Qualifications
   3. Filling Knowledge Gaps
   4. Guiding the Typology Development
   5. Directing Future Research

Introduction

As noted in Chapter 4, through an extensive review of existing data sets, a data set was identified with potential for informing the development of a typology of homeless families. The Fragile Families and Child Well-Being Study (Reichman, N.E., Teitler, J.O., Garfinkel, I., McLanahan, S., 2001) follows a birth cohort of new parents and their children over a 5-year period beginning in 1998. This sample is at high risk for homelessness in that pregnancy is one of the major risk factors found to precede homelessness (Weitzman, 1989) or loss of housing (Rog and Gutman, 1997). In addition, because the study oversampled unmarried women, the sample contains a higher proportion of women potentially more vulnerable to residential instability. Furthermore, Fragile Families is a national longitudinal panel study that includes measures of residential instability and risk (e.g., being evicted, having utilities turned off), as well as the incidence of being doubled-up (i.e., staying for at least one night with others) and literal homelessness (i.e., staying at least one night in a literally homeless situation).⁽¹⁾ These data thus afford the ability to track families over time and to examine the role of various risk or protective factors on residential stability.

This chapter describes a reanalysis of the Fragile Families data set focused on the following research questions:

• What are the risk and protective factors that differentiate, among a cohort of poor families, those families who:
  • Experienced homelessness and those who remained stably housed?
  • Experienced homelessness and those who become doubled up or residentially unstable (i.e., at risk of homelessness)?
  • Became doubled up or residentially unstable and those who remained stably housed?

The reanalysis is intended to inform our conceptualization of a typology of homeless families. As a multisite database of high-risk families, it provides an opportunity to examine the incidence of homelessness over multiple geographic areas, over time, and in contrast to a comparison population of poor families experiencing a range of residential arrangements.⁽²⁾

The chapter begins with a brief description of the data set, the sample selected for the re-analyses, the creation of residential groups and other relevant measures, and the analyses performed. Then, the results of the reanalysis are provided, followed by a summary of the findings and a discussion of the study's implications for filling knowledge gaps, guiding typology development, and directing future research.

[ Go to Contents ]

Methodology

Database Description

The Fragile Families and Child Well-being Study, also referred to as the "The Survey of New Parents," is designed to track a cohort of new parents and their children over a 5-year period. The purpose of the study is to provide new information on the strengths, conditions, and relationships of both wed and unwed parents and how Federal and state policies affect family composition and child well-being. The study is a stratified random sample of U.S. cities with a population of 200,000 or more, designed to provide a representative sample of nonmarital births in U.S. cities with populations over 200,000. Mothers were approached and interviewed at the hospital within 48 hours of giving birth, and fathers were interviewed at the hospital or elsewhere as soon as possible after the birth.

Eventually, four waves of data will be available. Baseline data were collected between 1998 and 2000, and followup interviews were conducted at 1 year, 3 years, and 5 years after the baseline. Currently, three waves are available for reanalysis: baseline, Year 1 followup, and Year 3 followup. Baseline data are available on a sample of 4,898 families (3,712 nonmarital births and 1,186 marital births). One-year followup data are available on a total of 4,365 mothers and 3,367 fathers, and Year 3 followup interviews are available on 4,231 mothers and 3,299 fathers. At least one wave of followup is available on 94 percent of the mothers, while 82 percent of the mothers were interviewed at both followups.⁽³⁾ The database constructed for this reanalysis focuses on the mother, with data about the father given where appropriate.

Defining the Sample for Reanalysis

The Fragile Families data set includes families from diverse income backgrounds, ranging from those far beneath the income poverty level to those who have relatively high levels of income. For an analysis examining the risk factors for homelessness, it is important that the groups being compared have an equivalent probability of experiencing the condition. Therefore, an income limit (i.e., household income at or below 50% of national poverty) was used to define the sample. In addition, the sample was limited to families in which the mother was 18 years of age or older. Finally, the sample was selected based on the families completing the Year 1 interview (n = 4365) because residential information was only collected during the followup interviews. A total of 838 families (19.2% of the Year 1 Fragile Families data set) met the income and age criteria and constitute the primary sample used for this study.

Creating and Describing Residential Outcome Groups

Detailed residential information was collected on participants in the Fragile Families study at the Year 1 and Year 3 followup surveys. This residential information found in each survey included:

• # — Moves: Number of moves since birth of child/last interview;
• Residential Risk Indicators: Indicators of residential risk in past 12 months (i.e., had not paid full amount of rent or mortgage; had been evicted from home or apartment; had not paid full amount of a gas, oil, or electric bill; had phone service disconnected because payments were not made; had to borrow money from friends or family to help pay bills);
• Doubled-Up: Whether the family was currently living with family or friends and paying no rent, or had moved in with other people even for a little while due to financial problems in last 12 months; and
• Homeless: Whether the family was currently living on the street, in temporary housing or a group home, or spent at least one night in a shelter, abandoned building, automobile, or other place not meant for regular housing in the past 12 months.

For descriptive analyses, each mother was categorized into one of four residential groups at Year 1 and Year 3 based on the pattern of responses to these residential indicators (# moves, residential risk indicators, doubled-up, homeless).

As Table 5-1 shows, residentially stable households in Year 1 and Year 3 were defined as having less than two moves, no residential risks, and had not been doubled-up or homeless during the prior 12 months. At-Risk households reported two or more moves and/or one or more residential risks, and also had not been doubled-up or homeless in the last 12 months. Doubled-up households were ones that were currently or recently doubled-up and had not been homeless, regardless of the number of moves or residential risks they reported in the past 12 months. Homeless households were ones that reported currently living on the street, in temporary housing or group home, or had spent at least one night in the past 12 months in a shelter, abandoned building, automobile or other place not meant for regular housing.

In addition to categorizing households into one of these four residential groups at Year 1 and Year 3, a combined residential group was also created based on the most severe residential category experienced in the two waves. A family who was residentially stable during Year 1 but doubled-up at Year 3, for example, would be classified as doubled-up. In order to be considered residentially stable, a family would need to meet the stable criteria for both Year 1 and Year 3. Conversely, to be put into the homeless group a family only had to report being homeless in Year 1 or Year 3.

Potential Risk and Protective Factors

Variables to be examined were selected in part based on characteristics that were found to be important in past research along with those proposed by members of an Expert Panel, convened to guide the conceptualization of the typology (a detailed meeting summary is included in Chapter 3). Demographic and background variables were examined, including the mother's age, race, and whether her first birth was as a teenager, as were several household characteristics, such as the number of children in the household and whether the mother was living with her spouse/partner, living with her own mother, or living with other adults (not including her spouse/partner). Variables were also examined that allow us to describe services used by these households, including receipt of health services, employment training, child care, and housing-related services, such as living in public housing or receiving housing assistance. Changes in health status, alcohol and substance use, and mental health have also been examined. Reports of whether the mother had recently been hit or slapped by her partner/spouse/child's father were also combined to create a measure of domestic violence. Table 3-1 provides a complete list of the variables that were examined.

Descriptive analyses, described below, were conducted with all of the variables shown in Table 5-1. Only those variables that showed substantial variation between housing groups (e.g., there were statistically significant differences between stably housed or homeless and at least two of the three other groups) or were considered important background and demographic characteristics, however, were included in the multivariate analyses.

[ Go to Contents ]

Descriptive Analyses and Results

Descriptive analyses were conducted with all of the variables shown in Table 5-2. These analyses were conducted to examine differences among the combined residential groups on the range of variables listed in Table 5-2. Alcohol and drug use were combined to create a single substance use variable. Several measures of mental health status — currently feeling sad or depressed, recently lost interest in hobbies/work, or recently feeling tense/anxious — were also combined into a single mental health indicator. Means, standard deviations, percentages and other descriptive statistics were computed for these variables for each residential group. The appropriate comparative analysis — chi-square, t-test — was then used to determine if the residential groups statistically differed on each of these variables. These analyses allow us to determine how these individual groups compared and contrasted.

The number of families in each residential group varies, both over time and when combined. At Year 1 and Year 3, for example, over one-third of families can be classified as residentially stable (35% and 42%) or at-risk (39% and 37%), while approximately one-fifth were doubled-up (21% and 16%) and only one-twentieth were currently or recently homeless (6% and 5%). When the two years are combined, however, the number of residentially stable families declines to only 22%. The percentage of families at-risk across both time periods increases to 41%, and Doubled-Up to 28%. The percentage of families ever homeless also increases to 8%.

The tables found in Appendix E provide the descriptive comparisons of the four combined residential groups for households at or below 50 percent of the poverty level on all key variables. The table also shows statistical differences between and among the groups.

This section provides a brief summary of these findings, highlighting the patterns of differences among the groups. First, as has been found in past studies (see Chapter 2), there were no demographic and background differences between the various residential groups. The mother's age at baseline, for example, was almost identical between the four groups (24 to 25 years old, on average), and comparable percentages of women were African-American (61% - 70%). There were also no major differences in the percentage of women with a high school degree (45% - 49%), currently attending any school or training (14% - 20%), or working (34% - 38%).

There were, however, distinguishing characteristics for each of the groups. Stable families were most distinct from all other families on a full host of health, mental health and substance use variables. Compared to each of the other groups, families who were residentially stable both years reported statistically:

• Better health
• Less alcohol use
• Less drug use
• Less smoking
• Less daily interference from drug and alcohol use
• Less depression or other mental health issues
• Less likelihood of being hit or slapped by a spouse/partner

The other area of pronounced difference between stable families and all other groups of families involved resources. Of all four groups, those stably housed were most likely to have a spouse working and have someone who could co-sign for a loan. They were least likely of all groups to receive food stamps or free food in the past year, to report going hungry, and to apply for the Earned Income Tax credit. Although there are other differences between the residentially stable groups and others, the patterns of resources and problems are the strongest and most consistent.

Among the groups, at-risk families were the least likely to have lived with their mother at any interview time point, had the fewest number of adults in the family, and were most likely to have received a housing subsidy since the baseline. Doubled-up families, not surprisingly, are the most likely of all groups to have more adults in their household. Compared to at-risk families, doubled-up families are more likely to live with their mother, less likely to have a spouse or partner working, but more likely to have another adult working in the household, and less likely to have a housing subsidy.

Homeless families, compared to all groups, are most likely to have received free food in the past year, yet also most likely to have gone hungry, least likely to have someone in their family offer a place to live or to have someone who could co-sign a loan, and most likely to report using drugs and report mental health symptoms.

These descriptive comparisons show a variety of differences between the groups, but most clearly show differences between the groups on household composition, resources and receipt of benefits, and on health, mental health, and substance use.

[ Go to Contents ]

Predicting Residential Stability and Homelessness

A second set of analyses were performed to answer the questions:

• What are the risk and protective factors that differentiate homeless families from all others?
• What are the risk and protective factors that differentiate residentially stable families from all others?

To answer these questions, statistical procedures (logistic regressions) were used that could test for the effects of all relevant variables at one time (rather than one at a time, as in the descriptive analyses). By looking at all variables simultaneously, it is possible to identify variables that are relatively more important in distinguishing residentially stable families from all others or those that are relatively more important in distinguishing homeless families from all others. The variables that set residentially stable families apart from others may be considered "protective" factors for homelessness and residential risk, while the factors that distinguish homeless families from all others can be considered potential "risk" factors for homelessness.

Logistic regressions were computed for Year 1 groups, Year 3 groups, and the combined residential groups. Only those variables that showed substantial variation between housing groups (e.g., there were statistically significant differences between stably housed or homeless and at least two of the three other groups) or were considered important background and demographic characteristics were included in these analyses. Each logistic model began by entering all of the variables in the model, and then removing non-significant variables⁽⁴⁾. Tables 5-4 and 5-5, which show the results from these logistic analyses, list all of the variables that were initially included in the model (e.g., non-shaded variables), but parameter estimates are only shown for those variables that were statistically significant at the .05 level in the final models.

Three models examined the factors that related to a family experiencing recent homelessness at Year 1, Year 3, and at either time-point. Three additional models examined the factors that related to a family remaining residentially stable at Year 1, Year 3, and at both time-points.

Homelessness. Table 5-3 presents the results of the three homeless models (Year 1, Year 1 and 3, and Year 3). The Nagelkerke R2 (a pseudo- R2 statistic that measures the amount of variance explained by the model) for the Year 1 and Year 1-3 models are both less than .2, indicating that neither model is doing a very good job of fitting the data. The Year 3 model has a Nagelkerke R2 of .333, however, indicating that this is a better fitting, more powerful model (closer to Cohen's definition of a medium effect).

Only one variable, income, is significant in all three models. Families with relatively higher household incomes were consistently less likely to experience homelessness, an effect that was strongest for the Year 3 model (parameter estimate of -.303).

A few variables were significant in two of the three models. Receiving housing assistance (local, state, or Federal) appears to be a protective factor. People who reported receiving housing assistance at baseline or Year 1, as well as those who obtained housing assistance during the followup period, (having a negative coefficient for the change score) were also less likely to experience homelessness.

Mental health issues, substance abuse issues, and reports of domestic violence were also somewhat related to a greater likelihood of experiencing homelessness. Finally, receipt of TANF was positively related to the likelihood of becoming homeless, but was likely a proxy for need and lack of income rather than a predictor of homelessness.

Stably Housed. The descriptive analyses showed that it was often the Stably Housed group that differed the most from the other residential groups. Table 5-4 presents models that examine factors to predict who was residentially stable at Year 1, at Year 3, as well as Year 1 AND Year 3. The overall fit of all three models is fairly consistent and low; Nagelkerke R² of .221 for the Year 1 model, .183 for the Year 3 model, and .197 for the Year 1-3 model (all would be considered small effects). Table 5-4 presents the results for the stably housed group.

Looking for results that were significant in more than one model showed that living with a partner/spouse, at least at baseline, made it more likely that a mother would be residentially stable. Changes in this relationship, however, had contradictory effects in different models; in the Year 1 model, having a partner join the household was associated with greater likelihood of being stable, whereas in Year 3, the household was less stable if a partner joined (or more stable if the partner left). Perhaps this was due, in part, to the decrease in partner employment noted earlier in Year 3.

The more adults there are living in the household, and having a spouse/partner who is working or who has found employment, all make it more likely that a mother will be residentially stable. Living in public housing was also frequently associated with being stably housed, while obtaining public housing was significant only for the combined Year 1/Year 3 outcome.

Factors that made it less likely that someone would be residentially stable somewhat mirror the results of the homeless analyses. Reported substance use and mental health issues made it less likely that a woman would be residentially stable in all three models. Poorer reported physical health was also associated with a decreased risk of residential stability in the combined model, and reported domestic violence was significant in two of the models.

[ Go to Contents ]

Discussion

Summary of Results

The reanalysis of the Fragile Families database shows that even among women who are extremely poor (at or below 50% of the poverty level), the risk of being homeless is not very large. Using a very broad definition of homelessness, less than one in ten (8%) of the women in this poverty sample indicated that they had been homeless for even 1 night over a 1-to-3 year period. However, only 22 percent reported being residentially stable (moving no more than once, not reporting any problems making ends meet) for the entire period, while the largest group (40%) of women were generally residentially stable but experienced some sort of financial issues (e.g., had problems paying for food, housing and/or utilities), but had not been homeless, doubled-up, or had to move frequently.⁽⁶⁾

The Fragile Families reanalysis also shows that there are characteristics and experiences that distinguish between these residential outcomes. Bivariate analyses indicate that the residential groups are distinct on a number of variables, often in a linear fashion, from those who experience the least stability to those experiencing the most stability. The most consistent findings relate to mothers' health, mental health, and substance use, suggesting that these conditions heighten their vulnerability to become homeless and their absence helps a mother remain stable. Overall, however, the results of the logistic models find few variables that have strong predictive value in differentiating those who experience homelessness from all others living in vast poverty, or those who remain residentially stable from all others. Having higher incomes and receiving housing assistance appear to serve as protective factors in the homeless models, whereas the health, mental health, and substance use issues appear to place a mother at risk (though the findings are not entirely consistent). In predicting stability living with a partner relates to greater stability, especially if the partner is working. Having other adults in the household also appears to increase a mother's likelihood of remaining stable and, not surprisingly, having substance use and mental health issues lessens a mother's likelihood of remaining stable.

Caveats and Qualifications

Several important qualifications need to be kept in mind when reviewing all of these findings. One issue is the relatively small number of households in this poverty sample that were ever homeless during the period examined (less than 100). The small number of cases limits how much can be said even descriptively about these families. In addition, little information was obtained on the homeless experience. Thus, the group could include families who spent one night in shelter to those who spent many nights and had multiple episodes of homelessness.

For the logistic regression models, the relatively poor fit of most of the models (with Nagelkerke R² scores typically only around .2) should serve as a reminder to treat these findings with some caution. Although it is plausible that the low fit for the various homeless models could be attributed to the small number of cases in the condition or to the heterogeneous nature of the outcome variable, the fact that low model fits were found with the stably housed models where the numbers were greater and the definition of stable more solid makes this explanation less likely. It is more likely that the poor fit of these models is due to the reliance on individual-level variables and the absence of any contextual variables.

These issues notwithstanding, though, the reanalysis of the Fragile Families database has still provided an opportunity to address some of our knowledge gaps with respect to homeless families, guide our conceptualization of a typology, and inform designs for future research.

Filling Knowledge Gaps

Although not designed to provide information on homeless families, the Fragile Families database has provided information that is useful in filling some of our knowledge gaps with respect to homeless families. One important gap that this data set helps fill is providing information on a national sample of homeless families, rather than being restricted to a single city. In fact, looking more closely at the geographic location of families (e.g., were homeless families more likely than others to come from some metropolitan areas?), might be another useful analysis. Unfortunately, geographic data were not readily available to those who had access to the public data sets. Reingold and Fertig, in "The Characteristics and Causes of Homelessness Among At Risk families with Children in Twenty American Cities" included as Appendix D, had unrestricted access to the Fragile Families data and did examine a few contextual variables. However, only the number of shelter beds in a city related to the probability of experiencing homelessness. It is possible, however, that unexplored contextual variables may be important to examine in predicting not only homelessness but residential stability as well.

The Fragile Families data set is also useful in that it provides information on a broader sample of at-risk families. As already noted, a key finding from this reanalysis is that there is a range of residential patterns experienced by even very poor families and that it is as useful to determine what keeps families stable as it is to know what predicts homelessness. This type of analysis is difficult to do with the typical homeless database but was possible in this reanalysis.

The fact that the Fragile Families project has collected information over time is also important, providing a longitudinal perspective that is often missing from studies. The longitudinal analyses not only showed that the incidence of homelessness was relatively rare (less than 10% ever homeless), even in this extreme poverty sample, it also showed that only a handful of households (9 households, 1% of the eligible families) reported being homeless in more than one time period. It is also true, however, that less than a quarter of the families remained stable throughout both time periods.

The eventual release of the 5-year followup survey should provide even more opportunities to examine the residential patterns of these various households, including a chance to examine households that fall back into homelessness, as well as what predicts long-term stability. The small number of families that experience homelessness, however, will likely make it difficult to do many analyses with such a group even if they could be identified. The small number also makes it difficult to use the Fragile Families data set to examine any subsets of homeless families, such as those who are working or two-parent families.

Guiding the Typology Development

The relatively poor fit of the logistic regression models, examining both homelessness and residential stability, limits how much guidance this reanalysis of the Fragile Families database can provide for developing a typology of homeless families. The results do suggest that mental health and substance use issues (and to a lesser degree, domestic violence) increase a family's vulnerability to homelessness and that the absence of these issues heightens a family's probability of remaining stable. Housing assistance (such as receiving a subsidy) and having more money, not surprisingly, help families avoid homelessness, as has been found in prior studies.

As noted, the relatively poor fit of these models suggests that individual-level characteristics such as these are not the only factors involved in predicting who will become homeless. For those who are struggling well below the poverty level, it is likely that contextual factors, such as the availability of affordable housing in an area, play an even more important role in determining the likelihood of becoming homeless or staying in stable housing.

Directing Future Research

As noted, several suggestions on how the Fragile Families data set could be used for future research include looking more closely at geographic differences, as well as taking advantage of the next wave of surveys. More broadly, this reanalysis has shown the utility of looking at a broader range of families that may be at risk of becoming homeless. While the factors associated with being residentially stable somewhat mirror the factors related to who becomes homeless, there are also important differences that can be seen only when it is possible to examine each group separately.

Additional new research may benefit from exploring more clearly the role that other family members (e.g., partner, other adults) play in fostering stability, as well as how the various health/mental health/substance use/domestic violence issues increase one's vulnerability. Do these issues make it difficult for a mother to work and thus rise out of poverty? Do they make her more vulnerable to being evicted or being thrown out of other relatives' homes? Do they make it difficult for other adults to remain living with them? Understanding the role these factors play may help in developing interventions that can prevent homelessness, especially among those who may have had and lost subsidies.

Although our analyses did not focus squarely on those living at risk or doubled-up, it is clear that these groups experience a number of stresses and their share of health, mental health, and related issues. Understanding their vulnerability and interventions that can help them rise to greater stability would be important to decreasing the daily challenges these families experience.

Lastly, studies should investigate how context interplays with individual-level factors and determine what community factors can play a role in fostering greater stability and decreasing the risk of homelessness.

[ Go to Contents ]

Endnotes

1. U.S. Department of Health and Human Services. For the purposes of some Federal definitions, being doubled up is considered homelessness whereas in other programs it is not.

2. Reingold and Fertig also conducted analyses on the Fragile Families data set for Appendix D in this volume (Reingold and Fertig, 2006) in response to a request to write a paper on at-risk families. Their paper was designed to examine homelessness among poor families with children while this chapter, derived from our exploration of relevant secondary data sets, looks more broadly at the different residential histories of poor women (50% or below the poverty level) and the factors that predict both homelessness and stability. Where the analyses overlap, similar results are found in both studies.

3. More detailed information on the Fragile Families data set can be found in Reichman et al. (Reichman, Teitler, Garfinkel, and McLanahan, 2001), as well as on the study's web site.

4. More specifically, the backward stepwise procedure removed non-significant variables one-by-one. Once all appropriate variables had been removed, however, the program re-examines all of the removed variables to see if any should be re-entered.

5. The outcome tables show all of the variables that were initially included in the model (nonshaded parameters), but parameter estimates are shown only for those variables used in the final model.

6. An interesting observation is that homelessness in this sample does not appear to be completely correlated with poverty. A total of 230 families (5% of the total sample) experienced homelessness at some point during the followup period; only one-third of these families were living at 50 percent of or below the poverty level and 29 percent were living above the poverty level. Additional analyses of these groups may provide further insights into the factors related to families becoming homeless.

Where to?

Top of Page | Contents
Main Page of Report | Contents of Report

Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)