Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Feb 15.
Published in final edited form as: J Fam Soc Work. 2018 Jun 18;21(4-5):331–347. doi: 10.1080/10522158.2018.1469563

The Impact of Housing Instability on Child Maltreatment: A Causal Investigation

Katherine E Marcal 1
PMCID: PMC6377199  NIHMSID: NIHMS984146  PMID: 30774282

Abstract

More than a half million children are confirmed as victims of maltreatment by the child welfare system each year. Children from unstably housed families are over-represented in child mal-treatment reports, and a growing body of evidence links housing problems to maltreatment and Child Protective Services (CPS); investigation. The present study applies two propensity score analysis approaches—greedy matching and propensity score weighting—to data from the Fragile Families and Child Well-being Study to move toward a causal explanation of child mal-treatment behaviors among mothers in low-income households. Utilizing two separate methods to correct for overt selection bias, the present study finds that housing instability leads to a small increase in maltreatment behaviors, yet this small positive net impact on child maltreatment does not fully explain the over-representation of unstably housed families in the child welfare system. Families experiencing housing problems likely have a range of needs that require earlier, targeted intervention to mitigate consequences of poverty, domestic violence, and maternal depression. Child welfare services should invest resources in housing assistance programs in-house as well as through partnerships with local public housing authorities to stabilize families, reduce housing-related strain on caregivers, and promote family preservation.

Keywords: Child welfare, families, housing instability and homelessness, maltreatment, propensity score analysis

Every year, more than a half million children are confirmed as victims of maltreatment (Anne E. Casey Foundation, 2017). Numerous socioeconomic risk factors including housing instability have been implicated, but causal pathways remain poorly understood. A growing body of evidence indicates housing problems elevate risk for maltreatment and Child Protective Services (CPS) investigation. Mothers facing extreme economic hardship are more likely to engage in harsh, aggressive, or abusive parenting practices than other mothers (Park, Ostler, & Fertig, 2015; Warren & Font, 2015), but the causes of maltreatment and appropriateness of child welfare responses are not well understood. The pathways to maltreatment and child welfare intervention remain murky, and decision makers must develop policy in the absence of empirical evidence. The present study applies two propensity score analyses to assess the net impact of housing instability on mothers’ maltreatment behaviors toward their children.

Background

In 2015, an estimated 683,000 children were victims of maltreatment in the United States; more than one third of these children were removed from their families by the child welfare system (U.S. Department of Health and Human Services, 2016, 2017). Maltreatment—defined as physical, sexual, or emotional abuse, or neglect—has detrimental effects across multiple domains of child and adolescent development. Child maltreatment is often an indicator that families are facing extremely stressful or high-conflict situations. Several factors that strain families have been linked to abuse and neglect. Unemployment, single parenthood, and earlier childbearing can all create stress in families that elevates the risk for maltreatment (Frioux et al., 2014; Runyan, Wattam, Ikeda, Hassan, & Ramiro, 2002). Unsurprisingly, children in poverty are especially vulnerable as families navigate constant material scarcity and face difficulties affording basic needs. The link between financial hardship and child maltreatment has been well documented (Cherry & Wang, 2016; Escaravage, 2014; Featherstone at al., 2017). Children living in families who earn incomes that fall below the federal poverty level (FPL) are more than three times as likely as nonpoor children to experience maltreatment (Drake & Jonson-Reid, 2014), and high levels of neighborhood poverty are associated with elevated rates of physical abuse, sexual abuse, and neglect (Drake & Pandey, 1996; Shuey & Leventhal, 2017). Caregivers in struggling households face stressful conditions such as chaotic or unsafe living environments along with inadequate resources. Conditions of scarcity and stress impact caregiver behaviors and resources available to families, leading to disparate levels of maltreatment and child welfare intervention (Featherstone et al., 2017).

Housing instability and maltreatment

Intrinsically linked to poverty, housing instability further strains families as caregivers—primarily single mothers—struggle to raise children in unstable or unaffordable conditions. Housing instability represents a uniquely distressing experience, particularly among families with children as they undergo school changes and disruption of social networks while adjusting to new environments. Caregiver stress is elevated among mothers experiencing housing instability (Park et al., 2015; Suglia, Duarte, & Sandel, 2011), and high stress contributes to risk for maltreatment (Berger, 2004; Warren & Font, 2015). Disproportionate representation of unstably housed families in the child welfare system suggests a connection between housing instability and maltreatment that threatens child well-being and overburdens service providers (Park, Metraux, Brodbar, & Culhane, 2004). An examination of the scope of housing problems among a nationally representative sample of child welfare–involved families found housing instability contributed to risk for foster care placement for one in six children under investigation for mal-treatment; moreover, housing problems delayed reunification for children already placed out of home (Fowler et al., 2013). Beyond the effects of general poverty, exposure to housing-related hardship appears to represent a unique threat to child safety by straining mothers and impeding healthy parenting environments, placing children at risk and leading to undue burden on child-serving agencies.

A large body of literature links various types of housing problems with increased risk for child maltreatment, although the majority of evidence is correlative (Slack, Berger, & Noyes, 2017). A county-level examination of child welfare cases across two decades in Pennsylvania yielded a positive association between foreclosure rates and volumes of investigated and substantiated child maltreatment (Frioux et al., 2014). Using a large secondary data set of children born in 20 large American cities, Park and colleagues (2015) found elevated likelihood of physical and psychological parenting aggression among homeless and doubled-up mothers compared to similarly low-income housed mothers. A recent study aiming to isolate the effects of inadequate housing found that housing instability and unaffordability independently increased the risk for maltreatment above and beyond poverty (Warren & Font, 2015). Further evidence suggests addressing housing instability may alleviate systemic challenges in child welfare and promote family preservation; in a study of homeless families with school-age children in Minnesota, receipt of supportive housing services was associated with a reduction in involvement with CPS across a 3-year period (Hong & Piescher, 2012). Despite a wealth of empirical support for the link between housing instability and child maltreatment, reliance on observational data leave the nature of this relationship unclear. It remains unknown whether inadequate housing itself leads to maltreatment, or whether families experiencing housing problems face unique challenges that place children in danger. This knowledge gap has important implications for developing and allocating services to alleviate and prevent maltreatment in vulnerable households.

Recent research points toward a causal link between poverty and child welfare involvement. Leveraging natural experiments made possible by policy shifts, researchers have been able to test the impacts of income changes on children and families. In one such study, low-income unmarried families with children in Wisconsin underwent randomization that led to exogenous differences in income; in the experimental condition, mothers received the full child support payments to which they were entitled whereas the control group received only partial or no child support (Cancian, Yang, & Slack, 2013). Families in the experimental group were significantly less likely to become involved in the child welfare system compared to control families. In a second study, policies that increased states’ minimum wages reduced the risk of children being reported for maltreatment (Raissian & Bullinger, 2017). Similarly, a study utilizing a large secondary data set of low-income unmarried families found that an exogenous increase in income led to a decrease in child neglect and child welfare system involvement (Berger, Font, Slack, & Waldfogel, 2016). This emerging body of research provides support to the theory a causal link exists between economic hardship and child welfare involvement, which may signal elevated incidence of maltreatment.

Given that family housing instability most often occurs in the context of poverty, it is plausible to expect a similar causal link may exist between housing instability and maltreatment. Despite ample evidence of correlation between housing problems and maltreatment and some evidence of a causal link between poverty and maltreatment, the net impact of housing on mal-treatment is not currently known. Studies frequently use child welfare involvement as a proxy for maltreatment; although useful as an indicator of risk, CPS investigation does not provide a complete picture of parenting environments and caregiver behaviors (Drake & Jonson-Reid, 2000). Systematic differences between unstably housed and similarly vulnerable low-income housed families may confound efforts to assess the true effect of housing conditions on family functioning; other factors that precede housing instability and maltreatment may differentiate families such that any attempts to isolate the true effect of housing conditions are hindered. Mothers in unstably housed or homeless families are more likely than their housed counterparts to be younger, African American, unmarried, not living with a partner, unemployed, experiencing extreme financial hardship, victims of domestic violence, suffering from depression, and have more children (Curtis, Corman, Noonan, & Reichman, 2013; Shinn et al., 1998). Therefore, thorough and accurate assessments of the impact of housing instability on maltreatment must correct for these group differences to obtain unbiased estimates.

Present study

The present study aimed to expand upon existing evidence that establishes an associative relationship between housing instability and mothers’ maltreatment behaviors. A large sample of at-risk families was leveraged and analyses were modeled on a program evaluation design; housing instability was conceptualized as a dichotomous “treatment” condition whereby those who experienced housing instability comprised the treatment group, and those who did not experience housing instability comprised the control group. Two propensity score analysis methods—greedy matching and propensity score weighting—were applied to balance families who did and did not experience housing instability on a number of characteristics to correct for overt selection bias. Outcome analyses estimated the effect of housing instability on mothers’ maltreatment behaviors after correcting for these confounders and assess the extent to which the net difference in maltreatment behaviors could be attributed to differential exposure to housing instability. Greater understanding of the pathways to maltreatment is required in order to identify families at greatest risk and bolster prevention efforts.

Methods

Data

Data for the present study came from the Fragile Families and Child Well-being Study (“Fragile Families”). Fragile Families was a longitudinal study that followed a cohort of nearly 5,000 children born 1998 to 2000 in 20 large American cities. Participants were selected according to a stratified clustered sampling strategy that oversampled children born to unmarried parents (“fragile families”). Mothers were interviewed in hospitals shortly after giving birth. Follow-up interviews occurred at 1-, 3-, 5-, and 9-year intervals. The present study utilized data from the mother interviews at the baseline as well as 5- and 9-year interviews. The sample was limited to mothers who reported having primary custody of their children and with complete data on all study variables (N = 2,284).

Measures

Dependent variable

The outcome of interest was the frequency of maltreatment behaviors committed by each mother toward the study focal child at the 9-year follow-up interview. Mothers’ maltreatment behaviors were assessed using a modified version of the Parent-Child Conflict Tactics Scale (CTSPC) available in the Fragile Families data set (Bendheim-Thoman Center for Research on Child Wellbeing, 2013; Straus, Hamby, Finkelhor, Moore, & Runyan, 1998). Fifteen of the original 22 CTSPC items were asked in the Fragile Families survey; items about severe physical abuse were excluded, whereas supplementary questions on neglect were added. Mothers self-reported whether they had completed certain disciplinary acts toward their children such as hitting, name-calling, cursing, threatening, or types of neglect such as whether they had had to leave the child home alone or were unable to care for the child due to substance use. Although self-report introduces potential social desirability bias, the CTSPC has been found to show agreement between self-reported and third party-reported maltreatment behaviors (Lee, Lansford, Pettit, Bates, & Dodge, 2012). Furthermore, the CTSPC has been validated in nationally representative samples of American parents (Straus et al., 1998) as well as samples from the Fragile Families data set (Taylor, Guterman, Lee, & Rathouz, 2009). A total maltreatment score was derived as the sum of positive responses; higher values indicated greater incidences of maternal maltreatment behaviors. The maximum possible score was 19 maltreatment behaviors.

Independent variable

The main predictor was a dichotomous measure of housing instability at the Year 5 interview. Mothers reported whether they had difficulty affording rent, mortgage, or utility payments; had been evicted for nonpayment of rent; had moved in with friends or family members to avoid becoming homeless; or spent time living in a homeless shelter or on the street. Although no standard definition of housing instability exists, similar indicators have been utilized in other studies of vulnerable families to capture a range of housing-related risk (Marcal, 2018; Ma, Gee, & Kushel, 2008; Park et al., 2004).

Matching variables

The first set of covariates were used to obtain propensity scores and balance the sample such that no systematic differences existed between families who did and did not experience housing instability at the Year 5 interview. All covariates used for matching were obtained at the Year 5 interview unless otherwise indicated. Mother’s age at the time of the child’s birth was measured in years and assessed at baseline. Race, also assessed at baseline, was categorized as White, Black, Hispanic, and other. Household income at Year 5 was calculated as a percentage of the FPL and categorized as below 100%, 100% to 299%, and 300% or more. Additionally, mothers indicated whether they were receiving welfare cash assistance, Social Security benefits, or Supplemental Nutrition Assistance Program (SNAP) benefits. Mothers reported whether they were married to or cohabitating with a partner (0 = no, 1 = yes), and whether they were employed (0 = no, 1 = yes). Based on the skewed distribution of household sizes, number of children in the household was dichotomized (0 = one or two children, 1 = more than two children).

Maternal depression was assessed using the World Health Organization’s Composite International Diagnostic Interview-Short Form (CIDI-SF; Kessler, Andrews, Mroczek, Ustun, & Wittchen, 1998). Mothers answered a series of questions that indicated probable major depressive disorder according to a threshold consistent with the diagnostic criteria in the DSM-IV (American Psychological Association, 2000; 0 = not depressed, 1 = depressed). The scale shows strong correlation with the original CIDI-long form scale, which has established acceptable validity (Kessler, Andrews, Mroczek, Ustun, & Wittchen, 1998; Orlando et al., 2006; Patten, Brandon-Christie, Devji, & Sedmak, 2000). Domestic violence was a dichotomous indicator of whether mothers reported having been hit, kicked, or slapped by the focal child’s father or a current partner in the past year.

Predictors

A second set of covariates was used as control variables in regression analyses testing the impact of housing instability on child maltreatment. Mother’s age, race, Year 9 income level, and Year 9 marital/cohabitating status were included in regression. Additionally, parenting stress indicated the degree to which mothers felt overburdened by child care obligations at Year 9. The measure was based on items used to develop a measure of aggravation in parenting for the Job Opportunities and Basic Skills Training Child Outcomes Study (JOBS; Child Trends Inc., 1993). The scale used in the present study included four items that asked about perceived stresses associated with parenting such as “Being a parent is harder than I thought” and “I feel trapped by my responsibilities as a parent” with Likert-type response options (1 = strongly agree, 4 = strongly disagree). Responses across the four items were averaged such that scores ranged from 1 to 4, with higher scores indicating higher levels of parenting stress. The scale had an alpha of 0.62 among mothers in the Fragile Families, which is considered acceptable in scales with fewer than 10 items (Bendheim-Thoman Center for Research on Child Wellbeing, 2013; Pallant, 2007).

Analytic approach

The goal of the present study was to determine the effect of housing instability on maltreatment behaviors among vulnerable mothers. Given the observational nature of the data, corrective methods were required to address the problem of selection bias and assess net effects. Two types of propensity score analysis methods were applied: greedy matching based on propensity scores and propensity score weighting.

Obtaining propensity scores

By correcting for bias in how participants are predisposed to fall into certain groups, propensity score analyses improve internal validity in assessing the true net effect of the treatment on a particular outcome. The propensity score is used to balance the treatment and control groups on key covariates such that no systematic observed between-group differences exist. Propensity scores are obtained by conducting a logistic regression with the dichotomous treatment condition as the dependent variable and key covariates drawn from prior literature as predictors; this generates predicted probabilities of selection into the treatment group for each participant based on a combination of characteristics. The logit of the predicted probability is defined as the propensity score (Equation 1).

q^(x)=log[1−e^(x))/e^(x)] 1
Greedy matching.

A subset of propensity score analysis methods uses matching—treatment cases are matched to control cases based on propensity scores. One of the most popular types is one-to-one nearest neighbor within-caliper greedy matching. In this procedure, each treatment case is matched to exactly one control case. Matches are selected based on the smallest distance between propensity scores within a defined range (a caliper). A recommended caliper size is 0.25 times the Standard Deviation of the propensity score (Guo & Fraser, 2015; Equation 2).

|Pi−Pj| < ε,j∈I0 2

where ε = 0.25σp

After matching is performed, the researcher is left with a new sample that includes only the matched cases; therefore, a weakness of this approach is that the sample size may be greatly reduced if many cases cannot be matched. However, a major strength is that systematic differences between the treatment and control groups are reduced, increasing internal validity in outcome analyses. Bivariate analyses should be used to assess the effectiveness of the matching procedure on balancing the sample. t Tests and chi-squared tests examine the differences on key covariates by treatment condition before and after matching; significant tests indicate the presence of selection bias. Effective matching should remove most significant differences between groups. After sufficient balance has been achieved, outcome analysis can be conducted on the matched sample. The present study matched the sample on mother’s age, race, household income level, employment status, marital status, family size, receipt of welfare, receipt of Social Security, receipt of SNAP, experiences of depression, and exposure to domestic violence. After matching and subsequent balance checks, linear regression was conducted on the matched sample in order to test the effect of housing instability on maltreatment behaviors.

Propensity score weighting

Given that there are multiple corrective methods to address the problem of selection bias, one method may be used as a sensitivity analysis for another. The current study applied propensity score weighting (PSW) as a check on the findings of greedy matching. PSW does not rely on matching procedures and thus serves as a useful robustness check of findings from methods that do match cases (Guo & Fraser, 2015). Furthermore, PSW maintains the original sample size, thus avoiding the major weakness of greedy matching. Propensity scores generated from logistic regression (described above) are used to calculate sampling weights that reflect the likelihood of a case being in the treatment group (Equation 3, where W indicates treatment status: 1 = treated, 0 = control; Guo & Fraser, 2015).

ω(W,x)=We^(x)+1−W1−e^(x) 3

Propensity scores were used to calculate a weight for each mother based on her propensity to have experienced housing instability; separate weights were calculated for those who had (W = 1) and had not (W = 0) experienced housing instability. Cases were weighted in outcome analyses to estimate the average treatment effect (ATE) for treated and control cases.

Balance checks were performed by incorporating weights into a series of linear regression models using each matching variable as an outcome and the treatment variable (housing instability) as the sole predictor. Nonsignificant coefficients would indicate sufficient balance had been achieved through PSW. Finally, the outcome analysis consisted of a linear regression model using the independent variable housing instability along with covariates to predict child maltreatment while incorporating propensity score weights.

Results

Sample description

Descriptive statistics yielded a relatively low-socioeconomic status sample. One in five mothers reported that their families struggled with housing instability at the Year 9 interview. Mothers were on average just older than age 25 years (SD = 5.9) at the time they gave birth to the study focal child, and 44% had families with at least three children. Approximately one half (49.6%) of participants were Black. More than one in three households had total incomes that fell below the FPL at the Years 5 and 9 interviews (38.2% and 34.8%, respectively). The average parenting stress score mothers displayed at Year 9 was 2.1 on a 4-point scale (SD = .45). Bivariate analyses showed treatment and control groups differed to a statistically significant degree on age, race, marital and cohabitating status, and family size, indicating the presence of selection bias in the sample and justifying the use of corrective methods (Table 1).

Table 1.

Sample description and group differences before and after greedy matching on propensity scores.

Uncorrected Sample After Greedy Matching
Housing Instability Housing Instability
No Yes No Yes
(n = 1,837) (n = 447) t or χ2 (n = 424) (n = 424) t or χ2
Age
 M(SD) 25.29 (6.11) 23.89 (5.05) 4.45*** 23.69 (5.49) 24.11 (5.07) −1.15
Race
 White 85.21 14.79 9.65** 46.27 53.73 0.89
 Black 76.61 23.39 20.82*** 51.83 48.17 1.79
 Hispanic 83.21 16.79 3.66 48.59 51.41 0.18
 Other 84.42 15.58 0.80 38.89 61.11 0.91
Poverty Level
 Below poverty 72.59 27.41 55.04*** 50.45 49.55 0.08
 100%−299% 81.08 18.92 0.44 49.41 50.59 0.08
 300%+ 93.63 6.37 65.70*** 50 50 0.00
Employed 80.93 19.07 0.60 51.35 48.65 0.97
Receives welfare 66.91 33.09 57.13*** 49.57 50.43 0.02
Receives SSI 74.14 25.86 3.07 43.75 56.25 0.80
Receives food stamps 71.50 28.50 82.47*** 50 50 0.00
Three or more children 78.42 21.58 4.56* 49.51 50.49 0.08
Married or cohabitating 83.32 16.68 20.91*** 48.65 51.35 0.68
Victim of domestic violence 54.84 45.16 26.51*** 49.81 50.19 0.01
Depression 58.67 41.33 134.99*** 39.29 60.71 1.33
Open in a new tab
*

p < .05

**

p < .01

***

p < .001.

Note. SSI = Supplemental Security Income.

Matching results and balance check

A logistic regression model predicting housing instability yielded predicted probabilities used to calculate propensity scores. Examination of the distributions of the propensity scores by group showed a sufficiently large common support region for matching. Each treatment case was matched to a control case with the most similar propensity score within the designated caliper size. After matching, the sample size was reduced from 2,284 to 848 (424 treatment cases and 424 control cases). A postmatching balance check showed that all significant differences by group were removed. Thus, matching achieved good balance in the data (Table 1).

Regression results

Two multivariate linear regression models using the ordinary least squares method of estimation were conducted to assess the impact of housing instability on maltreatment behaviors (Table 3). The first model used the original, uncorrected sample (N = 2,284) and found that housing instability increased the expected number of maltreatment behaviors by 0.99 behaviors (p < .001). Younger mother’s age, race, and stress were all also significantly associated with increased maltreatment behaviors.

Table 3.

Regression models estimating the impact of housing instability on maltreatment.

Uncorrected Greedy Matched Propensity Score-Weighted
(N = 2,284) (N = 848) (N = 2,284)
b SE b SE b SE
Intercept 7.75*** 0.42 8.75*** 0.69 8.28*** 0.62
Age −0.05*** 0.01 −0.05** 0.02 −0.05** 0.02
Race
 Black 0.36* 0.16 0.07 0.28 0.35 0.23
 Hispanic −0.61** 0.18 −0.56 0.33 −0.53* 0,26
 Other −0.06 0.35 0.28 0.72 −0.11 0.40
Poverty level
 100%−299% 0.18 0.14 −0.04 0.20 0.02 0.19
 300%+ 0.33 0.19 0.21 0.35 0.23 0.27
 Married/cohabitating −0.24 0.13 −0.47* 0.20 −0.30 0.19
 Parenting stress 0.67*** 0.13 0.52* 0.21 0.48* 0.21
 Housing instability 0.99*** 0.15 0.65** 0.20 0.80*** 0.17
Model fit
 F 18.05*** 4.48*** 7.93***
 R2 0.06 0.04 0.06
Open in a new tab
*

p < .05

**

p < .01

***

p < .001.

The second model conducted the same linear regression as the first using the postgreedy matching sample. Results showed the magnitude of the effect of housing instability of maltreatment was reduced after matching. Exposure to housing instability increased maltreatment by 0.65 behaviors (p < .001) in the matched sample. Also significant were age, marital/cohabitating status, and stress.

Sensitivity analysis

A second corrective method was applied to the data as check of the robustness of the findings. The same propensity scores obtained above were used to calculate sampling weights to balance the groups of families who had and had not experienced housing instability on key covariates. A balance check conducted through a series of simple linear and logistic regression models indicated that weighting procedures substantially reduced overt selection bias in the sample; though nearly all covariates were significantly associated with housing instability in the uncorrected models, none remained so in the models that incorporated propensity score weights (Table 2). The lack of significant findings from the balance check indicated adequate balance was achieved through weighting.

Table 2.

Balance check before and after propensity score weighting.

After Propensity
Uncorrected Sample Score Weighting
(N = 2,284) (N = 2,284)
b SE b SE
Age −1.39*** 0.31 −0.21 0.34
Race
 White −0.42** 0.14 −0.26 0.17
 Black 0.49*** 0.11 0.21 0.13
 Hispanic −0.24 0.13 −0.04 0.15
 Other −0.28 0.32 −0,15 0.40
Poverty level
 Below poverty 0.78*** 0.11 0.07 0.12
 100%−299% −0.07 0.11 −0.00 0.13
 300%+ −1.48*** 0.20 −0.10 0.21
Employed −0.08 0.11 0.00 0.13
Receives welfare 0.91*** 0.12 0.01 0.13
Receives SSI 0.38 0.22 0.15 0.28
Receives food stamps 0.96*** 0.11 0.07 0.12
Three or more children 0.23* 0.11 0.04 0.13
Married/cohabitating −0.49*** 0.11 −0.20 0.13
Domestic violence 1.27*** 0.26 0.14 0.32
Depression 1.36*** 0.12 −0.02 0.13
Open in a new tab
*

Note. Balance check conducted through simple linear/logistic regression models using the treatment variable (housing instability) as the sole predictor variable; significant beta coefficients indicate imbalance between treated/control groups on covariates. SSI = Supplemental Security Income.p < .05

**

p < .01

***

p < .001.

To test the validity of the findings from the main outcome analyses, propensity score weights were then incorporated into the multivariate linear regression model conducted on the uncorrected and greedy-matched samples (Table 3). Results indicated housing instability continued to have a small but significant effect on child maltreatment; being unstably housed increased a mother’s predicted maltreatment behaviors by 0.80 behaviors (p < .001).

Discussion

Results of the present study show exposure to housing instability leads to a statistically significant but small increase in maltreatment behaviors among mothers of elementary school-age children not placed out of home. Findings are consistent across two types of models in terms of effect size and direction after correcting for observed confounders, indicating validity and robustness. Mothers in intact, inadequately housed families commit on average less than one more maltreating behavior toward their children than mothers in stably housed families.

Despite statistical significance, the clinical significance of the increase in mal-treatment behaviors observed in the present study is questionable. Multiple models show housing instability is associated with an increase of less than one maternal maltreating behavior, suggesting practical differences between mothers in stably and unstably housed families may be negligible. Accounting for overt selection bias, the present study indicates that despite the unique stressors facing inadequately housed mothers, inadequate housing in itself does not substantially increase the number of mothers’ maltreating behaviors. Findings thus point to issues in the processes of detecting, reporting, and intervening on child maltreatment. Although housing problems only account for small net increases in mal-treatment, inadequately housed families are disproportionately likely to become involved in the child welfare system (Fowler et al., 2013; Park et al., 2004).

There are a number of reasons inadequately housed families may be more likely to come into contact with the child welfare system despite only marginal effects of housing instability on maltreatment. Families without stable housing may be exposed to situations that increase their visibility to service providers and thus their likelihood of being reported for child maltreatment, such as entering homeless shelters or other housing programs (Park et al., 2004; Park, Metraux, Culhane, & Mandell, 2012). Furthermore, inadequately housed families may face stigma that leads others to perceive their parenting practices as unacceptable or incompetent (Cosgrove & Flynn, 2005; Rogers, Bobich, & Heppell, 2016). In what Park and colleagues (2004) call the “fishbowl effect,” the heightened scrutiny that comes with homeless service use leads poor families to become more likely to come under child welfare investigation the longer they remain in homeless services. Finally, children in these families may face abusive or neglectful treatment from family members; the chaotic living situations experienced by unstably housed children may expose them to many different caregivers, particularly if they experience frequent residential moves, which may heighten their risk for maltreatment. Although mothers in families struggling with housing problems may not display substantially more maltreating behaviors than similarly low-income housed families, they nonetheless likely have a myriad of other needs that contribute to overall vulnerability and capture the attention of service providers.

The present study builds on prior evidence of a correlational relationship between housing problems and child maltreatment and supports recent findings that indicate economic hardship leads to a small net increase in maltreatment behaviors among vulnerable caregivers. However, the current over-representation of inadequately housed families in the child welfare system appears to represent a misalignment of service allocation and need. The systematic differences between families with and without housing problems observed before corrective measures were applied suggest variable levels of need; inadequately housed families experience high rates of poverty, domestic violence, and maternal depression, necessitating services that address these vulnerabilities before a crisis occurs that results in child welfare investigation or the removal of a child from the home. Earlier intervention and more targeted service delivery for unstably housed families may be a viable strategy to promote family preservation and reduce excess burden on child welfare services. In line with prior research on this population, younger mothers face unique risks as they struggle to take on new adult and parenting responsibilities (Frioux et al., 2014). Furthermore, maternal stress has an independent net effect on maltreatment that is similar in magnitude to that of housing instability, suggesting that parents who struggle to manage parenting responsibilities are more likely to engage in abusive or neglectful behaviors toward their children. Parenting stress should be evaluated in addition to socioeconomic or other household stressors to identify families at greatest risk.

Some findings must be interpreted with caution due to limitations in the data and study design. Although corrective methods to address selection bias improve internal validity, external validity is often compromised. In the present study, greedy matching substantially reduces sample size. This sacrifices information from the data and undermines the generalizability of the findings. Additionally, Fragile Families only samples urban households and findings cannot be generalized to rural communities, which also struggle with inadequate housing and child maltreatment (Carpenter-Song, Ferron, & Kobylenski, 2016; Silovsky et al., 2011). Future studies should examine the effect of housing instability on maltreatment behaviors in the unique contexts facing rural households. Maltreatment behaviors were collected through self-report, which could allow for social desirability bias; the developer of the maltreatment scale used in the present study noted that scores should be considered lower bounds of actual maltreatment incidence for this reason (Straus et al., 1998). More detailed information about the home environment such as child maltreatment from other sources such as other caregivers or grandparents was not included, which may have affected mothers’ parenting practices.

Findings from the present study inform efforts to prevent and reduce maltreatment. Financial strain, particularly related to housing, should be targeted by policies and agencies aiming to promote child well-being. Housing instability signals high-risk families with diverse needs. Thus, policies and programs designed to promote healthy parenting should be expanded to incorporate a wider range of services such as domestic violence resources, mental health treatment, and affordable childcare. To ensure child welfare investigation and intervention are appropriate to families’ needs and avoid the “fishbowl effect,” service providers across child- and family-serving agencies such as homeless shelters, welfare agencies, and community health centers should screen for sociodemo-graphic needs such as housing instability, parenting stress, and maltreatment risk.

Probing the link between housing instability and maltreatment has important implications for understanding why maltreatment occurs. Existing interventions often emphasize parent training, behavioral skills training, and anger management; findings of the present study suggest financial hardship should be an important focus as well, and parents experiencing housing-related strain should be targeted for interventions that address diverse needs to prevent child welfare involvement. This knowledge can be leveraged to implement preventative interventions that reduce strain on vulnerable care-givers and the child welfare system.

References

  1. American Psychological Association. (2000). Diagnostic and Statistical Manual of Mental Disorders, 4th Edition Washington, DC. [Google Scholar]
  2. Anne E Casey Foundation. (2017). Children who are confirmed by Child Protective Services as victims of maltreatment. Retrieved from http://datacenter.kidscount.org/data/tables/6221-children-who-are-confirmed-by-child-protective-services-as-victims-of-maltreatment#detailed/1/any/false/573,869,36,868,867/any/12943,12942
  3. Bendheim-Thoman Center for Research on Child Wellbeing. (2013). Scales Documentation and Question Sources for the Nine-Year Wave of the Fragile Families and Child Wellbeing Study. Retrieved from https://fragilefamilies.princeton.edu/sites/fragilefamilies/files/ff_scales9.pdf
  4. Berger LM (2004). Income, family structure, and child maltreatment risk. Children and Youth Services Review, 26, 725–748. doi: 10.1016/j.childyouth.2004.02.017 [DOI] [Google Scholar]
  5. Berger LM, Font SA, Slack KS, & Waldfogel J (2016). Income and child maltreatment in unmarried families: Evidence from the Earned Income Tax Credit. Review of Economics of the Household, Online first. doi: 10.1007/s11150-016-9346-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cancian M, Yang M, & Slack KS (2013). The effect of additional child support income on the risk of child maltreatment. Social Service Review, 83, 417–437. doi: 10.1086/671929 [DOI] [Google Scholar]
  7. Carpenter-Song E, Ferron J, & Kobylenski S (2016). Social exclusion and survival for families facing homelessness in rural New England. Journal of Social Distress and the Homeless, 25, 41–52. doi: 10.1080/10530789.2016.1138603 [DOI] [Google Scholar]
  8. Cherry R, & Wang C (2016). The link between male employment and child maltreatment in the U.S., 2000–2012. Children and Youth Services Review, 66, 117–122. doi: 10.1016/j.childyouth.2016.05.008 [DOI] [Google Scholar]
  9. Child Trends Inc. (1993). The JOBS Child Outcomes Study: Overview Briefing. Retrieved from https://www.childtrends.org/wp-content/uploads/2013/01/JOBS-Child-Outcomes-Study.pdf
  10. Cosgrove L, & Flynn C (2005). Marginalized mothers: Parenting without a home. Analyses of Social Issues and Public Policy, 5(1), 127–143. doi: 10.1111/asap.2005.5.issue-1 [DOI] [Google Scholar]
  11. Curtis MA, Corman H, Noonan K, & Reichman NE (2013). Life shocks and homelessness. Demography, 50, 2227–2253. doi: 10.1007/s13524-013-0230-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Drake B, & Jonson-Reid M (2000). Substantiation and early decision points in public child welfare: A conceptual reconsideration. Child Maltreatment, 5, 227–235. doi: 10.1177/1077559500005003003 [DOI] [PubMed] [Google Scholar]
  13. Drake B, & Jonson-Reid M (2014). Poverty and child maltreatment Korbin JE & Krugman RD (Eds.), Handbook of Child Maltreatment (pp. 131–148). New York City, NY: Springer. [Google Scholar]
  14. Drake B, & Pandey S (1996). Understanding the relationship between neighborhood poverty and specific types of child maltreatment. Child Abuse & Neglect, 20, 1003–1018. doi: 10.1016/0145-2134(96)00091-9 [DOI] [PubMed] [Google Scholar]
  15. Escaravage JH (2014). Child maltreatment entrenched by poverty: How financial need is linked to poorer outcomes in family preservation. Child Welfare, 93(1), 79–98. [PubMed] [Google Scholar]
  16. Featherstone B, Morris K, Daniel B, Bywaters P, Brady G, Bunting L, & Mirza N (2017). Poverty, inequality, child abuse and neglect: Changing the conversation across the UK in child protection? Children and Youth Services Review. doi: 10.1016/j.childyouth.2017.06.009 [DOI] [Google Scholar]
  17. Fowler PJ, Henry DB, Schoeny M, Landsverk J, Chavira D, & Taylor JJ (2013). Inadequate housing among families under investigation for child abuse and neglect: Prevalence from a national probability sample. American Journal of Community Psychology, 52, 106–114. doi: 10.1007/s10464-013-9580-8 [DOI] [PubMed] [Google Scholar]
  18. Frioux S, Wood JN, Fakeye O, Luan X, Localio R, & Rubin DM (2014). Longitudinal association of county-level economic indicators and child maltreatment incidents. Maternal and Child Health Journal, 18, 2202–2208. doi: 10.1007/s10995-014-1469-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Guo S, & Fraser M (2015). Propensity score analysis: Statistical methods and applications (2nd ed.). Thousand Oaks, CA: Sage Publications Inc. [Google Scholar]
  20. Hong S, & Piescher K (2012). The role of supportive housing in homeless children’s well-being: An investigation of child welfare and educational outcomes. Children and Youth Services Review, 34, 1440–1447. doi: 10.1016/j.childyouth.2012.03.025 [DOI] [Google Scholar]
  21. Kessler RC, Andrews G, Mroczek D, Ustun B, & Wittchen H (1998). The World Health Organization Composite International Diagnostic Interview Short-Form (CIDI-SF). International Journal of Methods in Psychiatric Research, 7(4), 171–185. doi: 10.1002/(ISSN)1557-0657 [DOI] [Google Scholar]
  22. Lee SJ, Lansford JE, Pettit GS, Bates JE, Dodge KA (2012). Parental agreement of reporting parent to child aggression using the Conflict Tactics Scales. Child Abuse and Neglect, 36, 510–518. doi: 10.1016/j.chiabu.2012.04.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Ma CT, Gee L, & Kushel MB (2008). Associations between housing instability and food insecurity with health care access in low-income children. Ambulatory Pediatrics, 8, 51–57. doi: 10.1016/j.ambp.2007.08.004 [DOI] [PubMed] [Google Scholar]
  24. Marcal KE (2018). Timing of housing crises: Impacts on maternal depression. Social Work in Mental Health, 16, 266–283. doi: 10.1080/15332985.2017.1385565 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Orlando M, Burnam MA, Sherbourne CD, Morton SC, London AS, Hays RD, & Bing EG (2006). Brief screening of psychiatric disorders among a national sample of HIV-positive adults: Concordance between the Composite International Diagnostic Interview (CIDI) and the CIDI Short Form (CIDI-SF). International Journal of Methods in Psychiatric Research, 10, 97–107. doi: 10.1002/mpr.104 [DOI] [Google Scholar]
  26. Pallant JF (2007). SPSS Survival Manual: A Step by Step Guide to Data Analysis using SPSS for Windows (3rd ed.). New York: McGraw-Hill. [Google Scholar]
  27. Park JM, Metraux S, Brodbar G, & Culhane DP (2004). Child welfare involvement among children in homeless families. Child Welfare, 83(5), 423–436. [PubMed] [Google Scholar]
  28. Park JM, Metraux S, Culhane DP, & Mandell DS (2012). Homelessness and children’s use of mental health services: A population-based study. Children and Youth Services Review, 34, 261–265. doi: 10.1016/j.childyouth.2011.10.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Park JM, Ostler T, & Fertig A (2015). Physical and psychological aggression toward a child among homeless, doubled-up, and other low-income families. Journal of Social Service Research, 41, 413–423. doi: 10.1080/01488376.2015.1018660 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Patten SB, Brandon-Christie J, Devji J, & Sedmak B (2000). Performance of the composite international diagnostic interview short form for major depression in a community sample. Chronic Diseases in Canada, 21(2), 68–72. [PubMed] [Google Scholar]
  31. Raissian KM, & Bullinger LR (2017). Money matters: Does the minimum wage affect child maltreatment rates? Children and Youth Services Review, 72, 60–70. doi: 10.1016/j.childyouth.2016.09.033 [DOI] [Google Scholar]
  32. Rogers KC, Bobich M, & Heppell P (2016). Further consideration of systems, stigma, trauma, and access to care. Pragmatic Case Studies in Psychotherapy, 12(2), 139–157. doi: 10.14713/pcsp.v12i2.1970 [DOI] [Google Scholar]
  33. Runyan D, Wattam C, Ikeda R, Hassan F, & Ramiro L (2002). Child abuse and neglect by parents and caregivers In Krug E, Dahlberg LL, Mercy JA, Zwi AB, & Lozano R (eds), World report on violence and health. Geneva, Switzerland: World Health Organization. [Google Scholar]
  34. Shinn M, Weitzman BC, Stojanovic D, Kickman JR, Jimenez L, Duchon L, & Krantz DH (1998). Predictors of homelessness among families in New York City: From shelter request to housing stability. American Journal of Public Health, 88(11), 1651–1657. doi: 10.2105/AJPH.88.11.1651 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Shuey EA, & Leventhal T (2017). Pathways of risk and resilience between neighborhood socioeconomic conditions and parenting. Children and Youth Services Review, 72, 52–59. doi: 10.1016/j.childyouth.2016.09.031 [DOI] [Google Scholar]
  36. Silovsky JF, Bard D, Chaffin M, Hecht D, Burris L, Owora A, & Lutzker J (2011). Prevention of child maltreatment in high-risk rural families: A randomized clinical trial with child welfare outcomes. Children and Youth Services Review, 33, 1435–1444. doi: 10.1016/j.childyouth.2011.04.023 [DOI] [Google Scholar]
  37. Slack KS, Berger LM, & Noyes JL (2017). Introduction to the special issue on the economic causes and consequences of child maltreatment. Children and Youth Services Review, 72, 1–4. doi: 10.1016/j.childyouth.2016.11.013 [DOI] [Google Scholar]
  38. Straus MA, Hamby SL, Finkelhor D, Moore DW, & Runyan D (1998). Identification of child maltreatment with the Parent-Child Conflict-Tactics Scales: Development and psychometric data for a national sample of American parents. Child Abuse & Neglect, 22(4), 249–270. doi: 10.1016/S0145-2134(97)00174-9 [DOI] [PubMed] [Google Scholar]
  39. Suglia S, Duarte C, & Sandel M (2011). Housing quality, housing instability, and maternal mental health. Journal of Urban Health, 88, 1105–1116. doi: 10.1007/s11524-011-9587-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Taylor CA, Guterman NB, Lee SJ, & Rathouz PJ (2009). Intimate partner violence, maternal stress, nativity, and risk for maternal maltreatment of young children. American Journal of Public Health, 99, 175–183. doi: 10.2105/AJPH.2007.126722) [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. U.S. Department of Health and Human Services. (2016). The AFCARS Report: Preliminary FY 2015 Estimates as of June 2016. Retrieved from https://www.acf.hhs.gov/sites/default/files/cb/afcarsreport23.pdf
  42. U.S. Department of Health and Human Services. (2017). The AFCARS Report: Preliminary FY 2016 Estimates as of October 20, 2017. Retrieved from https://www.acf.hhs.gov/sites/default/files/cb/afcarsreport24.pdf
  43. Warren EJ, & Font SA (2015). Housing insecurity, maternal stress, and child maltreatment: An application of the family stress model. Social Service Review, 89, 9–39. doi: 10.1086/680043 [DOI] [Google Scholar]

RESOURCES