Poverty, Homelessness, and Family Break-Up

Abstract

This study examines the extent and correlates of family separations in families experiencing homelessness. Of 2,307 parents recruited in family shelters across 12 sites, a tenth were separated from partners and a quarter from one or more children. Additional separations before and after shelter entry and reasons, from parents’ perspectives, were documented in qualitative interviews with a subsample of 80 parents. Separations were associated with economic hardship, shelter conditions, and family characteristics.

Poverty and homelessness are associated with the break-up of families. A number of studies have documented that children in families who experience homelessness frequently become separated from their parents. In a national sample in 1996, Burt et al. (1999) found that three fifths of women served by homeless programs had children under 18, but only 65% of the mothers lived with any of their children. Park et al. (2004) found that 24% of over 8,000 children who entered shelter with a parent for the first time in New York City in 1996 received child welfare services within five years of shelter entry; three quarters of them after the family became homeless. Many smaller, local studies document associations of housing problems and homelessness with elevated rates of foster care placements and far higher rates of informal child separations unknown to child welfare authorities (for reviews see Barrow & Lawinski, 2009; Courtney, McMurty, & Zinn, 2004).

Fewer studies have sought to explain these separations. In studies of the general population, poverty is related to child maltreatment, especially neglect (Sedlak et al., 2010) and to “substandard” parenting (Berger, 2007). Among families experiencing homelessness, Park et al. (2004) found that recurrent and longer shelter episodes and domestic violence predicted child welfare services. They suggest that families in shelters are subject to stress and lack of privacy, and also heightened scrutiny: a “fishbowl effect” may lead staff to report to child protective services. Similarly McDaniel and Slack (2005) suggest that life events, such as a move, may make low-income parents more visible, leading to protective service reports. Cowal et al. (2002) compared mothers who entered shelter with continuously housed mothers using public assistance and found that drug abuse, domestic violence, and any institutional placement of the mother predicted separation for both groups, but homelessness was by far the strongest predictor. Barrow and Lawinsky (2009) found the same factors along with children’s needs were important in a sample of mothers experiencing homelessness, but described precarious housing as “a constant backdrop.” In the face of crises that came “in twos and threes,” mothers negotiated with fathers, maternal and paternal kin, and agencies to find “better choices among troubling alternatives” (pp. 166–167) for themselves and their children.

The current mixed-methods study uses survey data to document the extent of child separations in a large multi-site sample of 2,307 families recruited in homeless shelters and uses both quantitative and qualitative data to examine explanatory factors posited in the literature. In particular in the survey data we examine associations of separations with the parent’s prior homelessness, substance abuse, domestic violence, felony conviction (a proxy for institutional placement), and foster care placement in childhood, along with parent and child demographic characteristics. The quantitative data also allow examination of the extent to which separations vary by site and shelter, suggesting policy differences in the homeless service and child welfare systems that may affect separations. Qualitative interviews with a subsample of 80 families elucidate from parents’ perspectives how poverty, housing problems, and the homeless service system contribute to separations.

Although our primary focus is on child separation, we additionally examine the extent to which partners are separated from each other. Families experiencing homelessness are often headed by single parents (Rog & Buckner, 2007), but this is partly a consequence of shelter and housing program policies that exclude men (U.S. Conference of Mayors, 2006). Thus we examine how separations of parents from partners they consider a part of their family varies by site and shelter, and reasons families give in qualitative interviews.

Methods

Participants

The Family Options study enrolled 2,307 families with children fifteen years of age and under who had spent at least one week in one of 57 emergency shelters in 12 sites,¹ drawn from all regions of the United States and varying housing and labor markets, from September, 2010 to January, 2012. Families were recruited into an experiment in which they received priority access to housing and service interventions. Very few families (n = 13) declined to participate, although 183 who failed to pass eligibility screening for available interventions were not enrolled. (Common reasons for exclusion included insufficient income or lack of employment, family composition, size of available units, poor credit history, criminal convictions, lack of sobriety.) We interviewed one adult at study enrollment, prior to random assignment, giving preference to mothers in two-adult families, because when parents are separated, children more frequently stay with the mother.

The adult respondents were predominantly female (91.6%), with a median age of 29. Over a quarter (27.4%) had a spouse or partner with them in family shelter. A plurality (43.7%) had one child with them in shelter but 11.1% had four or more. In half of the families (49.9%), at least one child was under 3. Study families were 41% African American, 21% white, non-Hispanic, 20% Hispanic (all races), 7% Asian/Pacific Islander and 11% mixed or other (with Hispanics excluded from all remaining categories). Families were deeply poor – median annual household income was $7,440 – and many came from poverty: during childhood 15.9% of respondents had been homeless and 27.1% had lived in foster care, a group home, or an institution. Poverty was also longstanding: 62.8% had experienced a prior episode of homelessness and 84.6% had been doubled up (living in the same unit with another family) as an adult because they could not pay the rent (for details see Gubits, Spellman, Dunton, Brown, & Wood, 2013).

We conducted semi-structured qualitative interviews with a non-random subsample of 80 families – 77 mothers and 3 fathers – from four geographically dispersed sites (Alameda County, Connecticut, Kansas City, and Phoenix) an average of 6.4 months after random assignment. Demographic characteristics of the subsample were similar to those of the full sample (for detail see Mayberry, Shinn, Benton, & Wise, 2014).

Measures

The adult respondent (in the full sample) reported on all family members who were with her in shelter and also about spouses, partners, and minor children “who are part of the family but are not living with you right now in [shelter name].” Additional variables are shown in Table 1 and described in detail in Gubits et al. (2015).

Table 1.

Logistic regression predicting child separations from parent and family characteristics (N=5,165 children)

Variable	OR	95% CI
Male child	1.07	[0.90, 1.26]
Number of children in household	1.35	[1.24, 1.46]***
Parent age	0.93	[0.92, 0.95]***
Single parent	0.83	[0.64, 1.06]
Previously homeless	1.43	[1.11, 1.83]**
Alcohol abuse	1.13	[0.81, 1.57]
Drug abuse	1.27	[0.93, 1.73]
Foster or institutional care in childhood	1.13	[0.87, 1.46]
Prior felony conviction	1.84	[1.31, 2.58]***
Adult domestic violence experience	0.90	[0.70, 1.16]
Categorical variables	df	Chi-square
Child age	3	248.99***
Household income category	5	23.85***
Race/Ethnicity	4	11.29*
Site	11	22.57*

Note. OR = odds ratio; CI = confidence interval. 231 observations excluded due to missing values. Chi-square indicates joint significance of categorical variables in the full model. Standard errors adjusted for clustering of children in families.

^*p<.05,

^**p<.01,

^***p<.001.

The qualitative interviews covered family composition, housing decisions, family routines and rituals, and social supports. The family composition section that is the focus here asked the respondent about separations from children (for any reason) and from partners (if associated with housing or housing programs). Additional questions probed for reasons for separation and how it unfolded, how long the respondent expected the separation to last, whether the respondent had reunified with the child and on what that depended, and whether the respondent felt the separation was the best option for the child. Respondents who had never separated from a child were asked whether there was ever a time when they had considered doing so, and why. Interviews averaged about an hour, with interviews where respondents reported separations taking longer than others. Interviews were recorded and transcribed.

Analyses

Quantitative predictors of child separation (listed in Table 1) were analyzed in SAS using logistic regression at the level of the child, with standard errors corrected for clustering of children within families. Because we were interested in separations from partners only if they were related to housing, we did not examine individual level predictors but tested only whether the proportion of families with a spouse or partner living elsewhere at the time of study enrollment differed by site and shelter. Analyses for shelter included the 42 shelters with at least 20 enrollments.

Qualitative interview transcripts were analyzed using NVivo9. Research team members each read a subset of the interviews. The team then developed a thematic coding scheme inductively for each section of the interview. Next, two analysts refined the coding scheme for a specific section of the interview and examined inter-rater reliability. Reliability for existence of and reasons for separation for children (across 32 interviews) and partners (across 20 interviews) were kappa = .85 and .91 respectively. Discrepancies were resolved by consensus. One analyst completed the remaining coding, but both discussed difficult-to-classify cases. We coded all instances of separation of the respondent from children, including normative separations (e.g. due to custody after divorce). For partner separations, we considered only separations related to housing and housing programs.

Results

The quantitative interviews provide data on the extent of separations among families experiencing homelessness. In the full sample of families who had spent seven days in shelter, 10.1% of adult respondents reported that a spouse or partner was living elsewhere. Nearly a quarter (23.9%) had a minor child who was not in the shelter with the family (living with other relatives, friends, in foster care, or in other living situations). Only 0.7% of respondents reported that a child was in foster care. Rates of separation in the quantitative data for the qualitative subsample of 80 were similar (10% for partners, 25% for children).

The quantitative interviews also allow for the identification of adult and child characteristics associated with child separations. Table 1 shows the results of a logistic regression predicting child separation. Child age (categorical variable) was strongly associated with the likelihood of separation. A third (33.7%) of children age 13 to 17 were separated compared to 22.2% of children age 8 to 12, 13.4% of children age 3 to 7, and only 4.5% of children age 0 to 2. There was little variation in separations by child gender, with 15.5% of girls and 16.7% of boys being separated, and no interaction between child age and gender (p=.86). Younger parents and those with more children, previous experiences of homelessness, and prior felony convictions were more likely to be separated from their children. Race and income (categorical variables) also mattered. Households that reported less than $5,000 in annual income had 2.6 times higher odds of having a separated child compared to households with incomes of $25,000 or more. White non-Hispanic respondents had 1.5 times the odds of having a separated child compared to black non-Hispanic respondents, with no differences between black non-Hispanics and other groups. Interestingly, alcohol and drug abuse, domestic violence at any time in adulthood, and having been in foster care as a child were not predictive of separations.

Rates of child separations ranged across sites from 9% to 24% of children (leading to a significant site effect controlling for parent and child characteristics). Separations were most common (exceeding 20% of children, 39% of families) in Salt Lake City, and Baltimore, and least common (below 10% of children, 13% of families) in Boston and Connecticut. Child separation rates differed by shelter (F(1, 41) = 4262.90, p<.0001) without other controls. Partner separations were also associated with site (F(1, 11) = 70.60, p<.0001) and shelter (F(1, 41) = 119.53, p<.0001), with separations highest in Baltimore (24% of all families; 88% of those with a spouse or partner) and lowest in Honolulu (4% of all families and 5% of those with a spouse or partner). Spousal (but not child) separations were generally higher in the East than in the Midwest or West.

The qualitative interviews help to explain the circumstances of these separations and others that occurred before and after the survey and how separations were influenced by poverty, housing, and housing programs. Of the 80 participants, 43 (54%) reported 57 instances in which they had been separated from a total of 78 minor children; if a family separated from two or more children at the same time, under the same circumstances, we considered this one instance. If a family separated from the same child on two occasions, we coded each instance separately. Most separations occurred during periods of homelessness or housing instability.

Reasons for child separations

Table 2 displays reasons for separation coded into eight categories with combinations of reasons coded into the uppermost category (row) on the list, because we deemed this more central. Thus, for example, if a respondent attributed her inability to feed her children adequately to lack of money for food, we coded this as hardship rather than inability to parent to own or family standards, or if protective service removed a child upon a respondent’s arrest, we coded this under arrest rather than protective services. Most of the children stayed with their other parent or another relative during the separations, but we coded these separations as normative (six families and instances) only if they were unrelated to the other reasons on the list. Half of the non-normative separations (24 families and instances) were related to economic and housing hardship, or shelter.

Table 2.

Reasons for separation of parent from minor children (with multiple reasons for a single instance counted in uppermost category on the list)

Code	Definition:	Instances of Separation	Families Affected	Children Affected
Shelter	Issues related to entries or living in shelter, including conditions of shelter, shelter rules that separate families, not wanting the child to be exposed to shelter.	9	9	13
Arrest	Respondent was arrested.	8	6	10
Protective Services	Protective Services removed child for reason other than parental arrest.	6	3	10
Hardship	Respondent chose separation due to poverty, housing instability, unemployment, hunger, or inability to provide for child.	15	15	20
Inability to parent to own or family’s standards, reasons other than hardship	Respondent was unable to parent the child appropriately in her own judgment, or that of family members who intervened. Includes respondent’s substance abuse.	5	5	6
Child behavior	Child’s behavior was dangerous to him/herself or others, or otherwise unacceptable, and respondent was unable to address the behavior.	3	3	3
Child safety	Respondent chose separation due to unsafe living situation. Includes domestic and neighborhood violence.	5	4	6
Normative other parent or relative custody	Other parent or family member has custody of child by respondent or child choice or custody decision, unrelated to issues above.	6	6	9
Total	A child or family could have multiple instances	57	43	76

Hardship

Economic hardship unrelated to shelter was the most common reason for separations (15 families and instances). In most cases, the family was experiencing housing instability, living in motels or doubled up with others because they could not afford their own place, or moving from place to place. Respondents described wanting their child to have stability and a sense of normalcy. In four instances parents were unable to provide the children’s basic needs.

At the time I was pregnant, and we were living in motels. I found myself getting broke. We were eating fast foods. I got paid from my job and I called their dad, and I said, “[Ex-Partner], I love my boys, I know you love them too, but I need help right now.” We met and he took the boys… I didn’t have a refrigerator or nothing like that, so I don’t want my boys to - - it was beginning to be too much.

Shelter

Nine families separated from children either upon shelter entry or during a shelter stay (sometimes one occurring before the study), typically so that children could avoid exposure to shelter conditions:

…it took its toll on my children. They were going to sleep in class because of the babies waking up in the middle of the night at the shelter.

I was letting her grandmother take her out of the shelter because she was losing weight and she was getting bad. She hearty but she was getting bad like the other kids.

In two cases, the shelter could not accommodate all minor children. Nor would shelters typically take extended families. When a three-generational family was evicted the mother and grandmother each took a child so both could go to family shelters. In another instance (not in Table 2), an adult child, age 20, was excluded from a shelter but later rejoined his mother and siblings in housing.

Parenting or child safety

Several categories of reasons for separation reflect the parent’s inability to care adequately for the child or to maintain a safe environment. Six respondents (eight instances) were arrested (all before study entry), and five (five instances) were unable to parent according to their own standards, or those of their family, most commonly due to substance abuse. Several of these parents sought treatment, and some were reunited with children afterwards. In three cases, relatives took the child from the parent because of the parent’s youth or substance abuse. Four respondents (five instances) cited child safety due to the environment inside or outside of the household. Typically, separations coded under Safety were related to hardship and housing instability but safety was the proximal issue. For example, one respondent had to move somewhere she deemed unsafe after an eviction; another left an unsafe area and moved in with a violent boyfriend. Three families (six instances) were separated by a protective services agency due to parental substance abuse and neglect. In all but two other instances, both involving arrest, the respondent arranged for family members to take the child without formal agency involvement.

Child behavior

In three families (three instances), separations began because of children’s behavior. In two cases children reacted badly to a move away from relatives, and the respondent sent them to those relatives; in the third the child was picked up by police and sent, briefly, to a mental health facility.

Best Option

Although many parents who were separated from a child described a sense of loss and reported that the child missed them, 34 of the 43 parents considered the painful decision to be the best option among difficult choices because the child was stable and better provided for. Separations enabled the child to remain in a good school or to develop a bond with extended family.

As much as it hurt me to be separated from my daughter, you know, sometimes you have to make sacrifices. You have to put them first. You have to think about what’s best for them for that time until things get better or you figure something out.

Other children had negative experiences, including one who was molested and another who was physically abused during the separation.

No Separations from Children

Among parents who had not been separated from children, 13 of 37 had considered a separation, typically for reasons associated with hardship or shelter:

Maybe before we got into the shelter, because it was hard to get into that shelter… So instead of taking my kids to a park, there was numerous shelters I had called. And there was one shelter where the woman said, “we can take your kids for the night so they don’t have to sleep outside. We’ll take them, but we can’t take you.” And I was like, well, if I have to sleep in a park, my kids are definitely going to go there. ‘Cause I don’t want them to do it. But – so yeah. I considered it then, but it didn’t happen. Thank god!

Partner Separations

Of the 80 respondents, 12 (14 instances) had been separated from a partner for housing-related reasons. Half of the partner separations had to do with rules of shelters or housing programs that excluded men, unmarried couples, or people with criminal convictions. Although respondents felt they had no housing options that would allow partners to stay together, many of them described the resulting strain:

[T]hen I had to move all the stuff out, and there wasn’t no help at the time, because it was just a shelter for women and children. He wasn’t with me … so it was like – if he was here, it would be so much easier, but they didn’t allow that.

Two respondents left doubled-up situations that had accommodated partners to obtain housing that they deemed better for themselves and their children. Two others moved into doubled-up situations that required separating from partners. One of these families was living in their car:

And it was just better for him (partner) to send me back to my family for me to get a support down there than it was for me to stay here. Because everybody was telling us that Children and Youth would come take my daughter if they found us in a car and all this. So we just wasn’t willing to risk that. So he just--we just sold the furniture and stuff that we had, and he bought us tickets and sent us back home.

Most respondents who experienced a housing-related separation from their partners reported negative impacts on their children.

…when [partner] did move in with us the baby was kind of like he knew who he was but it was kind of like hmm, I haven’t seen this guy in a while. Like where’d you come from? …. He didn’t really know who [partner] was and then he finally figured, oh yeah, this is my dad so he’s supposed to be around me.

Reunification with Children and Partners

All but five respondents anticipated that the separations from their children would be temporary. However, 20 of 57 separations (35%) lasted longer than the parent anticipated, often because it took her longer than expected to secure stable housing or to become financially able to care for the children.

Of the 57 incidences of separation from children, 34 had ended in reunification at the time of the qualitative interview. Respondents indicated that securing adequate housing permitted 14 of these reunifications. Nine parents reported that ongoing separations would continue until the parents secured housing. These parents were living in shelter or transitional housing (five), doubled up with other households in the same apartment (three), or in a subsidized apartment that was too small to accommodate all children (one). Thus the ending of nearly half of all separations (23/57) depended on housing.

Similarly, nine of the 14 separations from a partner had ended in reunification at the time of the qualitative interview, typically because the respondent or the partner was able to secure housing that could accommodate the entire family. Reunification in three additional cases depended on housing.

No parent indicated that shelters or other housing services attempted to reunite them with their families. Rather, shelter and housing programs tended to consider only members present with the respondent in evaluating housing needs, resulting in assignment to places too small for the full family. Respondents also reported that staff in shelters and transitional housing threatened to involve protective services if parents did not comply with shelter rules (Mayberry et al., 2014), and this led to one removal (where the parent violated a shelter rule about substance use.)

Discussion

As in other studies in the literature, this study shows that separations from children are rampant in families who experience homelessness. In our large 12-site sample, nearly a quarter of families who had spent a week or more in shelter were living apart from one or more of their children, although fewer than one percent had a child in foster care. Including separations at other times, over half of the qualitative subsample had been separated. Other studies have found that both separations and foster care placements often increase in the months following shelter entry (Cowal et al., 2002; Park et al., 2004), so the numbers may continue to grow.

Family demographic characteristics were associated with the likelihood of separations. Older children are much more likely to be separated from their families, with children age 13 to 17 being at particularly high risk. Mothers may be more likely to keep younger children with them, with older children more likely to stay with other relatives so that they are not exposed to shelter conditions or can maintain continuity in schooling. Despite some shelters having policies excluding older male children, no evidence of an interaction effect between age and gender was found. Larger households also faced greater difficulty staying intact or reunifying, perhaps in part due to constraints on unit size. White families are likely to have more resources to stay out of shelter than families of color; those who nonetheless become homeless may be more troubled, leading to higher rate of separations.

Both the quantitative and qualitative data point to the importance of extremely low incomes and resulting hardship in tearing families apart. Parents faced agonizing choices between keeping children with them and protecting them from shelter conditions or providing for their welfare. As in the study by Barrow and Lawinski (2009), most separations involved parental agency in difficult circumstances, and most separations were arranged informally between parents and other relatives.

Parental behavior also mattered. Arrests and felony convictions were associated with separations in the qualitative and quantitative data respectively. Substance abuse, perhaps surprisingly given previous studies, figured only in the qualitative data, and having experienced domestic violence as an adult was not associated with separations, perhaps because of the long time frame. Relatives sometimes intervened when they thought the respondent was not parenting appropriately.

Local policies also influenced parental options and choices, as evidenced by the fact that rates of separation varied substantially by site and shelter. Partner separations contributed to, but did not fully explain, the lower numbers of two-parent families in the East (as has been found in other studies, c.f. Rog and Buckner, 2007). Although some shelter staff threatened to call protective services in order to induce compliance with rules and did so in one case, the additional visibility of parenting under the watchful eyes of service providers does not explain informal separations. Children were rarely taken into foster care.

This study is the first to document the extent to which poverty and homelessness lead partners to separate from one another. One in ten parents had a partner living elsewhere while the family was in shelter. The quantitative and qualitative data clearly implicate shelters in separations of partners, although the fact that over a quarter of families in shelter had two parents suggests improvements over past years in shelters’ ability to accommodate at least nuclear families. Housing voucher programs also separate parents where one has a criminal record. The interviews show that the forced separation of fathers from their families is hard on mothers and children.

Implications for research and policy

We recruited families who had spent at least a week in shelter, and it is possible that families who can resolve homelessness quickly would have lower rates of separation than the families surveyed here. Nevertheless, results are troubling with implications for both research and policy. With respect to research, the fact that studies of children who experience homelessness exclude those who are separated from their parents means that samples are seriously biased. Whether child separations reflect hardship, parental behavior, or child behavior, children who are separated are likely to be faring worse than children who remain with their families. Estimates of effects of homelessness on children may be underestimates. Shelter policies excluding men may have led researchers to exaggerate the role of single parenthood in homelessness.

With respect to policy, programs that work with poor families, from income support and housing programs to shelters and transitional housing programs to correctional institutions to substance abuse treatment programs, should pay more attention to preserving families Separations are hard on both parents and children, and separation from parents in the family of origin is a predictor of future homelessness in adults (Rog and Buckner, 2007).

Family preservation may conflict with other policy goals. For example, in a congregate shelter or transitional housing program, one family’s husband and father may be seen as a potential danger to the next family’s child, and public housing rules designed to preserve the safety of the community by excluding criminals separate parents from their families. Welfare time limits may encourage adults to work, but lead to hardship, hunger, and ultimately separations for families. Prisons are designed to isolate and punish inmates – but the separation also punishes children and partners. Naming and quantifying the problem at least allows it to be taken into consideration in policy choices. Scatter-site homeless and housing programs, alternative sentencing, and substance abuse treatment programs that permit children to stay with parents may be able to reduce family separations. Housing programs should take family members living elsewhere into account in assigning units, to permit reunification. Child welfare authorities in particular should serve as advocates for minimizing separations of children from parents, and reunification when separations cannot be avoided.