Substitute care entry: The relationship between race or ethnicity and levels of county organization☆
Introduction
Subsequent to a maltreatment investigation, the decision to place a child in state custody hinges on several formal criteria. Child welfare workers must assess the level of harm a child has experienced, risk of future harm, likelihood that services provided at home will successfully minimize this risk, and any impact that placement in state custody would have for a child and his or her family (Shlonsky and Wagner, 2005, White and Walsh, 2006). Beyond these formal criteria, other factors may influence this decision. These factors may include a child's race, caregiver socioeconomic status, and structural characteristics of the community in which the decision occurs (Baumann et al., 2011, Shlonsky and Wagner, 2005). Although technically unrelated to formal decision criteria, these factors potentially influence the likelihood of entering substitute care and have been implicated in the overrepresentation of particular racial and economically disadvantaged groups in the child welfare system.
The disproportional representation of racial and economic minorities is well documented within U.S. child welfare jurisdictions. At every point in the decision making continuum, evidence suggests that racial minorities and low-income families are overrepresented. Children of racial and ethnic minorities and poor families are more likely to be subjects of a maltreatment investigation, enter substitute care, and stay longer in state custody than their White and more affluent counterparts (Fluke et al., 2003, Hill, 2007, Rolock et al., 2011, Wulczyn et al., 2005). One example of the complex relationship between child poverty, race, and child welfare involvement is found in the evolving debate about the relevance of poverty in child welfare outcomes and decision making. Recent findings from a nationally representative sample found that income was a stronger predictor of maltreatment than race; differences by race, while small for families with low socioeconomic status, were larger among families with higher socioeconomic status (Sedlak et al., 2010).
With respect to the decision to place children in state custody, an exact rendering of the variables, their interrelationships, and mechanisms that explain disproportionality remains elusive. Some studies suggest a direct and independent effect of race and socioeconomic status on this decision. Two studies examining race and family income reported independent effects of both factors (Lindsey, 1991, Lu et al., 2004). Though suggestive, these studies provide incomplete evidence, as neither included both race and socioeconomic status in its respective analysis. Research including both race and class has addressed this limitation and still observed independent effects of each variable. In an examination of a nationally representative sample, Hill (2005) demonstrated an association of race and socioeconomic status with entry in substitute care. Specifically, African American children and children of families on Medicaid were more likely than White children and non-Medicaid recipients to enter care. Additionally, in a California study, African American children referred to child welfare for maltreatment were more likely than other children to enter foster care after adjusting for allegation, age at referral, and number of siblings (Needell, Brookhart, & Lee, 2003). Further, the level of poverty in the zip code of the referral was positively associated with likelihood of placement in care, suggesting that community structural factors were related to child welfare decisions. The study, however, did not explore an interaction between race and zip code poverty.
Other research points to a complex relationship between race, class, and placement. Katz, Hampton, Newberger, Bowles, and Snyder (1986) found that the main effects for social class and race were not significantly related to placement in substitute care. There was, however, an interaction between social class and type of allegation such that Medicaid-eligible families were more likely than non-Medicaid-eligible families to experience the removal of a child in instances of physical abuse but less likely in instances of neglect (Katz et al., 1986). Although speculative, Katz et al. (1986) proposed that workers’ decisions about placement were sensitive to family economic context. The role of context was also explored in a Texas sample (Rivaux et al., 2008). Like previous studies, African American children and children from lower income families were more likely to enter state custody than their White and more affluent counterparts. Importantly, Rivaux et al. (2008) observed that, although African American children experienced an increased likelihood of substitute care, their cases received significantly lower risk of harm scores. Hence, disparate treatment manifested not because workers perceived more risk in cases involving African American children. Rather, if understood as a contextual element in the decision making process, race acted to shift workers’ risk tolerances. Rivaux et al. (2008) speculatively extended the boundaries of the importance of context from the family to the “cracked sidewalks and streets, broken and boarded windows on homes, graffiti, and groups of African American youth standing on corners” (p. 152), suggesting that these community level factors also influence workers’ tolerance for risk. Unfortunately, Rivaux et al. (2008) left the broader role of community conditions untested and focused exclusively on the family context in which child welfare decisions occur. Despite the research relating race and economic factors to the decision to place children in care, the evidence is not entirely conclusive.
Several studies observed no relationship between placement in substitute care and child or caregiver race or socioeconomic status, including one conducted in Illinois (Harris, Tittle, & Poertner, 2005) and two others from the Mid-Atlantic region (Runyan et al., 1981, Zuravin and DePanfilis, 1997). There are at least two potential reasons for these inconsistencies. Socioeconomic status is a notoriously difficult construct to operationalize (Krieger, Williams, & Moss, 1997). It encompasses wealth, income, professional and personal prestige, and legal status. Further, assessment of these dimensions may focus on different units of analysis such as the individual, household, or community. Developing indicators to tap all of these dimensions is onerous and few studies attempt comprehensive measurement of the construct (Krieger et al., 1997). Non-comprehensive indicators may only partially reveal the effect of socioeconomic status (Williams, Mohammed, Leavell, & Collins, 2010). The studies in our review relied on administrative data that contained limited socioeconomic information. As a consequence, these studies often used only one indicator of social class such as Medicaid eligibility, annual family income, or receipt of Aid to Families with Dependent Children.
It is also possible that the relationship of social class or race with child welfare involvement varies depending on other contextual elements. Work in decision making ecology postulates these elements moderate the effect of individual characteristics in decision making (Baumann et al., 2011), a point illustrated in a Canadian study. This study reported that racial and ethnic composition of agency caseload was associated with the likelihood of children of Aboriginal descent entering care (Fluke, Chabot, Fallon, MacLaurin, & Blackstock, 2010). Specifically, after covariate adjustment, the proportion of investigations involving Aboriginal caregivers was positively associated with likelihood that Aboriginal children would enter care.
Rivaux et al's. (2008) speculation finds support in literature examining the relationship between community context and maltreatment (e.g., Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007). In general, this research shows that certain community indicators are related to increased rates of maltreatment. In particular, as poverty, crime (Coulton, Korbin, Su, & Chow 1995), child care burden (Korbin, Coulton, Chard, Platt-Houston, & Su, 1998), community stability (Fromm, 2004), concentration of alcohol outlets (Freisthler, 2004), and vacant housing (Zuravin, 1989) increase so does reported maltreatment. Based on this research, one could view communities along a continuum of social and economic organization. Families living in communities at the highly organized end of the continuum would, on balance, experience relatively low levels of exposure to crime, impoverishment, child care burden, and residential instability. Conversely, families in communities at the other end of the continuum would experience greater exposure to these risk factors. The likelihood of maltreatment would track linearly with levels of community organization. The proposed mechanism to explain the relationship between maltreatment and community characteristics rests on at least two notions. First, compared to residents in highly organized communities, residents in socially disorganized communities are exposed to higher levels of chronic economic and emotional stressors that make maltreatment more likely. In addition to higher levels of ambient stress, these communities also tend to lack adequate social supports that protect against maltreatment. High residential mobility, for instance, prevents establishment and maintenance of positive social ties. High levels of poverty generate a dearth of material resources. As a consequence, maltreatment is more likely to occur in poor communities (Coulton et al., 2007). Additionally, these community characteristics tend to influence the definition, recognition, and the likelihood of reporting suspected maltreatment (Coulton et al., 2007). It is this early decision point where the largest racial and ethnic disparities in entry in state custody emerge (Needell et al., 2007, Rolock, 2011).
Recent research has deployed notions of community organization to test the relationship between community level characteristics and substitute care entry rates. In one large California County, Lery (2009) demonstrated that community levels of impoverishment, child care burden, and residential mobility predicted rates of entry into state custody. Other research has shown that community density of liquor stores is positively related to foster care removal rates (Freisthler, Gruenwald, Remer, Lery, & Needell, 2007). One multi-state analysis of child welfare data reported significant differences in rates of state custody between African American and White children and county level social indicators, reporting negative correlations between state custody and the county level percentage of African American residents, children living in poverty, female-headed households, and individuals without a high school diploma (Wulczyn & Lery, 2007). Although these relationships were unadjusted, the results suggest that county level characteristics may be related to the decision to place a child in state custody.
In this study, we build on past research through a simultaneous examination of the impact of individual and community factors on the likelihood of substitute care entry. We test the cumulative effect of community level factors and individual characteristics on the decision to place children in state custody. Like Wulczyn and Lery (2007), we chose county to represent community for two reasons. Given the relative novelty of this approach in child welfare, we sought a unit of analysis that could provide state-wide results, be useful and easily interpretable for child welfare professionals, and be replicated in other jurisdictions. In addition, county is an important political, organizational, and bureaucratic unit that is used to define boundaries for child welfare service units and the courts. As such, variation in placement decisions is likely to occur across counties. If effects present themselves at the county level, future research could opt for a finer grained analysis at the zip code or census tract. Finally, we chose to highlight the interaction between one individual level characteristic, race or ethnicity, with level of county social organization. As such, we test the combined effect of race or ethnicity and county level of organization on child welfare decisions.
Using individual child-level data for Illinois as well as county census, employment, and arrest data, we ask three questions: After an investigation of maltreatment, (1) is there a relationship between race or ethnicity and the likelihood of entering substitute care? (2) is there a relationship between community social organization as defined by socioeconomic, residential mobility, child care burden, and crime indicators and the likelihood of entering substitute care? and (3) is there a significant interaction between race or ethnicity and community level indicators on the likelihood of entering substitute care? Questions 1 and 2 replicate previous research. A tentative answer to question 3 offers new insights into the complex relationship between individual and community level factors on the decision to place children in state custody.
The following analytic plan describes a latent class analysis with twelve indicator variables of county level social organization. Selection of these variables was based on prior research establishing the importance of neighborhood characteristics for child maltreatment (Coulton et al., 1995), the relationship between structural characteristics of neighborhoods and criminal activities (Hipp, 2007), and the impact of neighborhood characteristics on rates of child maltreatment (Freisthler, 2004). Work by Coulton et al. (1995) and, more recently, by Lery (2009) used three factors to describe dimensions of social structural organization: impoverishment, residential instability, and child care burden, each constructed from a sub-set of variables. The child care burden factor, for instance, included the proportion of neighborhood residents older than 65 years as well as ratios of children to adults and adult males to adult females. The impoverishment factor included a number of variables such as child poverty rate, percent of community members who were African American, percent of female headed households, and unemployment rate. Some of these variables such as unemployment and child poverty are straightforward indicators of economic and material duress. Others such as the proportion of residents who are African American have a less straightforward association. We follow the lead of Coulton et al. (1995), who suggest that measuring the concentration of racial or ethnic minorities, African Americans in particular, is a way to assess indirectly “results of various types of disadvantage and discrimination” (Coulton et al., 1995, p. 1286). This does not suggest a causal relationship between race and poverty.
Section snippets
Analytic plan
The analysis consisted of two components. First, we conducted a latent profile analysis (LPA) to group counties in Illinois, using county level measures of community structure. We then constructed two logistic regression models to examine the relationship between entry into state custody and county level measures of community structure and individual demographic data on children investigated for maltreatment.
Using maximum likelihood estimation, the LPA classifies cases (counties) into a
Latent profile analysis
AIC, BIC, aBIC, and Log Likelihood values supported a model with 3 classes of county type. As seen in Table 1, the 3-class model had the lowest AIC and BIC values. An entropy value of .97 suggested minimal ambiguity in assignment of counties to their appropriate class (Celeux & Soromenho, 1996). A 4-class model failed to converge even after increasing the number of starting values. This suggested that a 4-class model was not identifiable (Muthén & Muthén, 2010a). Each county class fell loosely
Discussion
This study merges two distinct sets of research to understand the combined effects of individual level characteristics and community structural factors on the decision to place children in state custody. Using individual child welfare and county structural data, we posed three questions: After an investigation of maltreatment, (1) is there a relationship between race or ethnicity and the likelihood of entering substitute care? (2) is there a relationship between community levels of social
Limitations
The current study has limitations. While past research has used county as the unit of analysis (e.g., Aarons et al., 2010, Wulczyn and Lery, 2007), this level of analysis has some drawbacks. Smaller social and political units nest within counties. Contextual elements of these units may influence decision making. Aggregating to county blinds the analysis to these effects. Despite this shortcoming, county is a reasonable unit of analysis when examining the statewide behavior of a child welfare
References (52)
- et al.
Behavior problems and placement change in a national child welfare sample: A prospective study
Journal of the American Academy of Child and Adolescent Psychiatry
(2010) - et al.
How neighborhoods influence child maltreatment: A review of the literature and alternative pathways
Child Abuse & Neglect
(2007) - et al.
Latino children of immigrants in the child welfare system: Prevalence, characteristics, and risk
Children and Youth Services Review
(2009) - et al.
Disproportionate representation of race and ethnicity in child maltreatment: Investigation and victimization
Children and Youth Service Review
(2003) - et al.
Placement decisions and disparities among aboriginal groups: An application of the decision making ecology through multi-level analysis
Child Abuse & Neglect
(2010) A spatial analysis of social disorganization, alcohol access, and rates of child maltreatment in neighborhoods
Children and Youth Services Review
(2004)Neighborhood structure and foster care entry risk: The role of spatial scale in defining neighborhoods
Children and Youth Services Review
(2009)- et al.
Race, ethnicity, and case outcomes in child protective services
Children and Youth Services Review
(2004) - et al.
Black children and foster care placement in california
Children and Youth Services Review
(2003) - et al.
The next step: Integrating actuarial risk assessment and clinical judgment into an evidence-based practice framework in CPS case management
Children and Youth Services Review
(2005)
Convergence failures in logistic regression
SAS Global Forum 2008 Statistics and Data Analysis
The decision making ecology
An entropy criterion for assessing the number of clusters in a mixture model
Journal of Classification
Latent class and latent transition analysis
Community level factors and child maltreatment rates
Child Development
Exploring the spatial dynamics of alcohol outlets and child protective services referrals, substantiations, and foster care entries
Child Maltreatment
Illinois assisted housing and residential mobility guide: County level fact book and index
Factors that influence the decision to place a child in substitute care
The role of race in foster care placements
An analysis of racial/ethnic disproportionality and disparity at the national, state, and county levels
Income inequality, race and place. Does the distribution of race and class within neighborhoods affect crime rates?
Criminology
Returning children home: Clinical decision making in cases of child abuse and neglect
American Journal of Orthopsychiatry
Impoverishment and child maltreatment in African American and European American neighborhoods
Development and Psychopathology
Cited by (3)
Community perceptions and foster care placement: A multi-level analysis
2015, Children and Youth Services ReviewCitation Excerpt :Other research has shown that community density of liquor stores is positively related to foster care removal rates (Freisthler, Gruenewald, Remer, Lery, & Needell, 2007). A study in Illinois found that the level of social organization within a county was correlated with foster care entry, such that decreasing levels of social organization were associated with relatively higher odds of entry to care (Jantz, Rolock, Leathers, Dettlaff, & Gleeson, 2012). Testa and Furstenberg (2002) found that delinquent and neglected children were concentrated geographically and traced their presence to a limited set of Chicago neighborhoods that could be tracked back to the early 1900s.
The influence of differential response on decision-making in child protective service agencies
2015, Child Abuse and NeglectCitation Excerpt :Areas with concentrated poverty are more likely to have structural and social problems such as low-quality schools, high incidents of violence and criminal activities, few job opportunities, and high rates of social isolation (Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007). One study included poverty among several indicators of social disorganization and reported that children residing in counties with low levels of organization are more likely to enter substitute care during a maltreatment investigation (Jantz, Rolock, Leathers, Dettlaff, & Gleeson, 2012). Similarly, Drake and Pandey (1996) found that neglect, more than other forms of maltreatment, has a particularly strong association with neighborhood poverty.
What Will Happen to This Child If I Report? Outcomes of Reporting Child Maltreatment
2015, Child Maltreatment: Contemporary Issues in Research and Policy
- ☆
This study was funded in part through a grant from the Illinois Department of Children and Family Services.
- 1
The first two authors contributed equally to this study.