Learning Modules - Research and Evaluation - Center for Advanced Studies in Child Welfare - School of Social Work - College of Education and Human Development

Research & Evaluation

The Impact of Child Welfare Services on Racial Disparities in Outcomes: Child Welfare Services for African American and Caucasian Children in Four Minnesota Counties

Title IV-E Curriculum Module

Module Elements

Summary

The African American Disparities Committee of the Minnesota Department of Human Services, with assistance from the University of Minnesota conducted a case study that involved following neglect cases from the time of substantiation through a minimum of eighteen months or case closure (whichever came first). This study was conducted in an effort to improve methodologies used to examine possible racial disparities in the child welfare system. Using qualitative and quantitative analysis, answers to the following three research questions were sought: (1) What factors are associated with the selection and delivery of child welfare services to families? (2) Do services differ by race (specifically African American versus Caucasian) when controlling for other factors that impact service placement decisions? and (3) What factors, including race, are most influential in determining the placement of a child in out-of-home care?

Relationship to policy/practice

Findings from this study may help caseworkers understand sources of racial disparities in the child welfare system. Identifying sources of racial disparities allows policy makers and practitioners to isolate areas for improvement in the child welfare system.

Background

In the child welfare context, racial disproportionality is “the fact that some racial or ethnic groups of families and children are represented in various child welfare populations at levels that are disproportionate to their numbers in the overall family or child population” (Courtney & Skyles, 2003, p. 355-356). African American children are drastically overrepresented in the child welfare system. Although they only make up 15% of the general United States population, they make up 38 % of the child welfare population (BASSC, 2005).

Researchers have struggled to pinpoint the source of this disproportionality, especially since the National Incidence Study of Abuse and Neglect (NIS) has repeatedly found no differences in the incidence rates for child abuse and neglect for children of different racial/ethnic groups (Fluke, Yuan, Hedderson, & Curtis, 2003). In fact in its 1993 study, NIS found that when other risk factors were controlled, child maltreatment rates were significantly higher among Caucasian than African Americans (Hill, n.d.).

Courtney & Skyles (2003) propose that two independent mechanisms contribute to racial disproportionality. Children of a specific ethnic or racial group may enter the child welfare system at a rate disproportionate to their numbers in the population and/or they may exit the system at a slower rate than other racial or ethnic groups. The services associated with the former mechanism are termed “front-end services” and the latter can be termed “back-end services”.

In 2006, the Race Matters Consortium released an update on its Synthesis of Research on Disproportionality in Child Welfare (Hill, 2006). In this report, Hill (2006) reviewed studies from each decision-making stage of the child welfare process including: reporting, investigations, substantiation, placement in foster care, exit from foster care, and re-entry. Racial disproportionality findings or lack thereof were summarized for each stage of the child welfare process. A brief summary of Hill’s report and additional studies on racial disproportionality appear below.

Front-End Services

In confronting racial disproportionality, front-end services of the child welfare system have received considerable attention. Front-end services include referral/reporting, investigation, substantiation, and placement.

Some studies have found no racial disparities in child maltreatment reporting (See Hill, 2006). Other studies, however, have found grave disparities. For example, Ards, Myers, Malkis Sugrue, & Zhou (2003) found that in Minnesota, African American and Native Americans were six times more likely than Caucasian families to be reported for child maltreatment. Although not as strong, racial disparities in reporting have been found in other states such as California (Albert, 1994; Lu, Landsverk, Ellis-MacLeod, Newton, Ganger, & Johnson, 2004).

Research has revealed that hospitals under report maltreatment among Caucasians and over report maltreatment among African Americans (Hampton & Newberger, 1985). Roberts (2002) found that, for similar types of injuries to their children, African American families were more likely than White families to be reported. African Americans are more likely to be evaluated and reported for pediatric fractures due to alleged child abuse (Lane, Rubin, Monteith & Christian, 2002). Additionally, when an African American woman’s newborn tests positive for drug toxicity, she is more likely than a Caucasian woman to be reported for child maltreatment (Chasnoff, Landress & Barrett, 1990; Neuspiel et al., 1993).

According to research conducted by Fluke, Yuan, Hedderson, & Curtis (2003), African American children have disproportionately higher investigations of maltreatment than White children. When looking at racial disproportionality and the investigations process, some studies have found interactions between race and other factors. Gryzlak, Wells, and Johnson (2005) found an interaction between race and maltreatment type. Their research revealed that African families were more likely than Caucasian families to be screened in for investigation for neglect and physical abuse. Caucasian families were more likely to be screened in for investigation related to sexual abuse.

African American children also have an “elevated likelihood of substantiation compared to white children” (Ards, Myers, & Malkis, 2003 as cited in BASSC, 2005). Barth, Green and Miller (2001) revealed that African American children who receive public assistance were more likely to have an allegation substantiated. Thus, when looking at disproportionality and substantiation there may be interactions between variables such as race and socioeconomic status.

Due to confounding variables such as socioeconomic class, there have been inconsistent results with respect to studies that look at disproportionality and foster care placement. Some studies have not found race to be a predictor of foster care placement (See Runyan, Goud, Trost, & Loda, 1982; Katz, Hampton, Newberger, Bowles, & Snyder, 1986; Lindsay, 1994; Zuravin & DePanfilis, 1999; Harris, Tittle, & Poertner, 2001). But, a more recent study has found that children of color have a higher likelihood of being placed in foster care than Caucasian children (See Barth, Green & Miller, 2001).

Racial disparities in placement decisions have also been found. For example, in 1997 the U.S. Children’s Bureau reported that 56 % of African American children who were receiving child welfare services were placed in foster care; whereas, 72% of Caucasian children who were receiving child welfare services received services in the home. Research completed by Needell, Brookhart & Lee (2003) found that in California, even when other variables are taken into account, African American children were more likely than White or Latino children to be removed from their caregiver and placed in the child welfare system.

Back-End Services

Back-end services may include reunification, adoption, and re-entry. Harris and Courtney (2003) found that in comparison to being White or Latino, being African American in a single family was associated with a significant disadvantage with regards to reunification. They also found that in comparison to being White or African American, being Latino was associated with an advantage with respect to the timeliness of reunification. An evaluation of national data revealed that Caucasian children are four times more likely to be reunified with their families than African American children (Hill, 2001).

Not only are African American children less likely to be reunified, but they are also less likely to be adopted than their Caucasian counterparts (Courtney & Wong, 1996). Furthermore, researchers have found that the adoption process takes longer for African American children than for Caucasian children (Barth, Courtney, & Berry, 1994).

Researchers have consistently found no racial differences in re-entry rates. In 1999, Wulczyn looked at re-entry rates for five states. He found no significant differences between re-entry rates for African Americans and Caucasian children. In a study conducted in Oklahoma, Terling (1999) also did not find any significant differences by race in re-entry rates.

In sum, there have been inconsistent results when studying disproportionality and racial disparities in both front-end and back-end services. Methodological issues may explain the inconsistent results evident in these studies. Prior studies have relied heavily on the use of vignettes with workers and secondary analysis. These methodologies may be problematic. For example, vignette studies are particularly susceptible to the social desirability bias. That is, in an effort to please the researcher respondents may not respond to the vignettes in accordance with their actual behavior (Galante, 1999).

In relation to disproportionality studies, secondary analysis is problematic because the original research design is constructed for the purpose of tracking administrative case cost and process as opposed to analyzing race as a predictor in child welfare outcomes. Additionally, results from secondary analysis may be skewed due to the occurrence of the aggregation affect known as the Simpson’s Paradox. Generally in disproportionality studies, various race populations are combined into a single model where race is used as a variable in a larger regression. According to the Simpson’s Paradox aggregating subclasses (i.e. different non-majority races) may cause the relationship between two variables to disappear or even reverse (Curley & Browne, 2001, p. 309).

Project description, results summary

The racial disparities case record reading study was conducted by the Minnesota African American Disparities Committee, The Minnesota Department of Human Services (DHS), and the University of Minnesota School of Social Work. It was conducted in an effort to improve methodologies used to examine possible racial disparities in the child welfare system. The Minnesota Department of Human Services formed the African American Disparities Committee in response to the gross overrepresentation of African American children in Minnesota’s child welfare system. This committee, with assistance from the University of Minnesota conducted a case study that involved following neglect cases from the time of substantiation through a minimum of eighteen months or case closure (whichever came first).

In 2001, neglect cases from four Minnesota counties were identified. Neglect cases were chosen for this study because they are the main impetus behind African American children entering the child welfare system. To diminish the number of confounding variables additional selection requirements were established. The selection requirements included: children with single types of maltreatment; children who had not previously been placed; children who were only African American or only Caucasian; children who did not identify as having Latino ethnicity; children ages zero through eleven; and children whose parents were born in the United States.

One thousand ninety-five cases from the four counties met these requirements. Fifty-eight percent of these cases involved African American children and 48 % involved Caucasian children. The cases involving African American children were matched with cases involving Caucasian children on type of neglect, gender, age (0-5 and 6-11), and county. One hundred three pairs of cases were matched for a total of 206 children. Some cases were later eliminated because it was found that a child was of more than one race or that the parents were born in another country. The final sample was made up of 81 pairs totaling 162 children. The types of neglect cases included in the study were: abandonment (four cases), educational neglect (28 cases), endangerment (52 cases), inadequate supervision (14 cases), food/clothing/shelter neglect (58 cases), and prenatal exposure (six cases).

To review the cases a case record reading tool was developed that consisted mostly of checklist type question which lent themselves to a quantitative analysis along with some open-ended questions to provide data for a qualitative analysis. Case records were reviewed by teams of county supervisors and DHS personnel. Reviewers recorded the facts of the case for: investigation/assessment, case management in the home, or case management in out-of-home services (depending on which type of case management applied to the case). The case record reading tool was pre-tested in one of the counties six months prior to the case study and revisions were made accordingly. Finally, reviewer training sessions were held where reviewers were familiarized with the instrument.

Quantitative Findings

The McNemar chi-square statistic was used with cross-tabulations to determine if there were significant differences in case characteristics between the African American and Caucasian children when the pairs were not congruent. There was a significant difference in family composition at assessment. It was more likely that the father was not in the household when the child was African American. Additionally, the mothers of the African American children were more likely to be noted as being on public assistance and were more likely to be noted as being involved in domestic violence.

In cases where the pairs were not congruent, the worker was more likely to note an alcohol abuse problem when the child was Caucasian. When the child was Caucasian, it was also more likely that the mother was noted as having a physical disability, a cognitive disability, or a mental health problem. The police were more likely to arrange an informal placement for the Caucasian child when the pairs were incongruent. Finally, it was more likely for Caucasian children that the biological father, who was more often in the household, was a perpetrator.

There were no significant differences between the African American children and the Caucasian children with respect to parental drug abuse, inadequate housing, felony history, contact with the law, father’s disabilities, termination of parental rights for older siblings, or death of one or both parents. There were also no significant differences between the African American and the Caucasian child of the pair with respect to cases opened for case management services or cases sent to placement long enough to require reunification services.

Initially, a logistic regression of all 206 cases was conducted in order to better understand factors predicting ongoing placement (long enough to require reunification services). The dependent variable for the multivariate analysis was whether the child was in placement long enough (72 hours) to be referred for reunification services. Before the logistic regression was run, however, variables predicting placement and referral for reunification services needed to be identified. A qualitative analysis (which will be discussed later) was instrumental in deciding which variables to use in the logistic regression. In addition, the qualitative analysis aided in further reducing data anomalies and resulted in reduction of the sample of cases to 180 children who were not of mixed race or children of immigrant parents.

The probability that the case would be in placement long enough to require a referral for reunification services could be predicted from knowledge of the number of children in household, the interaction of worker and family, the age of the child, the history of maltreatment, the mother’s problems with the law and drugs, and whether substance abuse was extensive.

Regardless of race, the number of children in the household predicted whether or not the placement was long enough to require reunification services. The more children there were in the household the less likely the child would be in placement long enough to require reunification services.

Also regardless of race, the negativity of the worker and the cooperation of the parent predicted whether the case was referred for reunification services. When the worker was negative or the parent was uncooperative the case was 7.5 times more likely to go to placement long enough to require reunification services.

The logistic regression revealed an interaction between age and race. When a child was between one and four years old, the child was 1.2 times more likely to be in placement long enough to require reunification services. When the child was African American and between one and four years old, the child was less likely than the similarly aged Caucasian child to be in placement long enough to require reunification.

In order to reduce the number of maltreatment variables, a maltreatment summative scale from zero to three was created. Zero indicated no maltreatment history found at investigation, one indicated maltreatment history found at investigation, two indicated maltreatment was extensive or multigenerational, and three indicated maltreatment was both extensive and multi-generational. For every point higher the African American child was on the maltreatment scale, he or she was 2.9 times more likely than the Caucasian child to be in placement long enough to require reunification services.

Also, in order to reduce the number of variables with respect to the mother’s drug use and problems with the law, a drug and law problem scale was created. A zero on the scale meant the mother had no problems with drugs or the law. A one on the scale meant either the mother had problems with drugs or problems with the law. Finally a two on the scale meant the mother had problems with drugs and problems with the law. For every point higher on the scale a mother was, her child was 2.9 times more likely to be referred for reunification services regardless of race.

Extensive substance abuse was an additional predictor of whether or not the case was referred for reunification services. When there was a presence of extensive substance abuse (that is substance abuse that interfered with the care of the child), regardless of race, the case was 7.2 times more likely to be referred for reunification services.

Qualitative Findings

The purpose of the qualitative study was to clarify findings and take a deeper look at whether and how race affects case decisions and outcomes. In the qualitative analysis, data from the same 162 cases used in the quantitative analysis were used. The researchers re-read the original case record surveys that were completed for the quantitative analysis. Qualitative data based on the coding was entered into an electronic data base and the files were imported into NVIVO, a specialized software for qualitative analysis that finds patterns and trends in behavior.

From the qualitative analysis, child’s age emerged as being important in understanding differences among placement decisions and race. In the analysis, the following four age groups were established: less than one year old; one to four years old; five to nine years old; and ten or eleven years old. The themes and racial patterns that emerged in each age group will be briefly discussed.

Less than One Year Old

There were 30 children in the “children less than one year old” age group. These cases were largely positive toxicology and pre-natal cases. This group tended to have similar outcomes regardless of race. Consequently, these cases were omitted from further qualitative analysis.

One to Four Years Old

There were also 30 children in the “children one to four years old” age group. These cases could be divided into three categories. Thirty-three percent of these cases were parental competence cases, 54 % of the cases were crisis cases, and 13% of the cases were extensive substance abuse cases.

Two racial patterns emerged in this age group. Caucasian families were more likely to experience long term placement and to undergo termination of parental rights (TPR) or transfer of legal custody (TLC) than their African American counterparts. Racial patterns were also observed in whether the police made formal placements or allowed informal placements when placing the child outside of the home.

Of the 12 cases in which the police were involved, six involved Caucasian children. No placement was made in two of these cases, informal placement was allowed in two cases, and the police made a formal placement in two cases. For the six cases that involved African American children, only one case was not placed. The other five cases were formal placements. In other words, the police did not allow informal placements in any of the African American cases.

Five to Nine Years Old

Of the 162 cases analyzed, there were 74 children in the “children five to nine years old” age group. Thirty-four percent of these cases were parental competence and crisis case types, 40 % percent were educational neglect case types, 19% were parent not in home case types (due to police involvement or abandonment), and 7 % were parent or child mental health case types.

The qualitative analysis also uncovered racial patterns in this age group. First, African American children were more likely to enter long term placement in this age group. In other words, the racial trend found in the “children one to four years old” age group, for Caucasian children to experience long term placement, was reversed. Second, as with the “children one to four years old” age group, when placing children outside of the home the police were more likely to allow informal placements when the child was Caucasian. In fact, out of the five police involvement cases that were African American, the police only allowed one informal placement; whereas with the Caucasian cases the police allowed five out of five informal placements.

Ten to Eleven Years Old

There were 30 children in the “children ten to eleven years old” age group. Forty percent of these cases were educational neglect cases, 23 % where cases where parental drug or alcohol use was an issue, 17 % were parental competence case types, and 20 % were short-term crisis case types.

In this age group, African American and Caucasian children were equally unlikely to enter long term placement. Police only arranged out of home placements in three of these cases.

Conclusion and Recommendations

When a sample of African American children and Caucasian children were matched on type of neglect, gender, age group, and county, there were no significant differences in parental drug abuse, inadequate housing, felony history, contact with the law, the father’s disabilities, termination of parental rights for older siblings, or death of one or both parents. Notable differences (when pairs were not congruent) included: family composition; police involvement; whether worker noted alcohol abuse problem; whether the mother was noted to be on public assistance; whether the mother was noted to be involved in domestic violence; whether the mother was noted to have a physical disability, or cognitive or mental health problem, and whether the biological father was a perpetrator.

Surprisingly, a quantitative analysis revealed no statistically significant differences between African Americans and Caucasians with respect to percentage of cases opened for management services or the percentage of cases in placement long enough to require reunification services. Yet, racial disproportionality is overwhelmingly present in Minnesota. These findings suggest that, at least in Minnesota, racial disparities in the child welfare system cannot be understood without considering the interaction of race with other factors.

The multivariate analysis revealed that, regardless of race there are factors which predict whether or not a case will be in placement long enough to require reunification services. For example, regardless of whether a child is African American or Caucasian, the more children there are in the household the less likely a child will go into placement long enough to need reunification services. Worker negativity and lack of parental cooperation were additional factors that were predictive of placement regardless of race. Presence of worker negativity or lack of parental cooperation increased the odds that a child would go into placement long enough to require reunification services. Also, regardless of race, whether or not mom had a drug and/or law problem was predictive of being in placement long enough to require reunification services. Finally, the presence of extensive substance abuse predicted placement regardless of whether the child was African American or Caucasian.

Additionally, the multivariate analysis uncovered interactions between race and other factors such as age and maltreatment history. When the child was between the ages of one and four, the Caucasian child was more likely to be in placement long enough to need reunification services. When the child was between the ages of five and nine, this trend reversed; that is, within this age group African American children were more likely to be in placement long enough to need reunification services. Along those lines, when there was a history of maltreatment, African American children had greater odds of being in placement long enough to require reunification services in comparison to white children.

Selected Recommendations

Based on the above findings, several recommendations can be made. First, when looking at racial disparities future research should take into account the interaction of race with other case characteristics

Since African American cases with children of a certain age and with a maltreatment history are more likely to be in placement long enough to require reunification services, efforts to provide more appropriate services for these groups are warranted.

In addition, this study has identified factors that are influential in placement decisions regardless of race. In the future, researchers might explore why certain factors increase a child’s odds of being in placement long enough to receive reunification services. What service needs of these populations are not being met?