0009–4021/2008/0208151-168 $3.00 Child Welfare League of America 151
The Intersection of Race, Poverty,
and Risk: Understanding the
Decision to Provide Services to
Clients and to Remove Children
Stephanie L. Rivaux, Joyce James, Kim Wittenstrom,
Donald Baumann, Janess Sheets, Judith Henry, and
Studies have found that certain racial groups, particularly
the children of African American families, are placed in fos-
ter care at a higher rate than children of other races. These
families are also sometimes found to be afforded fewer
services that might prevent these removals, relative to fami-
lies of other races. It is unclear why this is so. Poverty has
been suspected, and sometimes found, to be the primary
cause of the disparity. Lacking in some of these analyses,
however, was how risk of future abuse/neglect to the child
entered into the decisions and particularly, how assump-
tions about race, poverty, and risk are factored into the
decision-making process. It is important to understand
this process if we are to find a way to correct it. The
current study addresses this process.
Stephanie L. Rivaux LMSW is Lecturer and Doctoral Candidate, The University of Texas
and Texas Department of Family and Protective Services, Austin, Texas. Joyce James
LMSW-AP is Assistant Commissioner for Child Protective Services, Texas Department of
Family and Protective Services, Austin, Texas. Kim Wittenstrom PhD is Research Analyst,
Texas Department of Family and Protective Services, Austin, Texas. Donald Baumann is
Head, CPS Evaluation, Texas Department of Family and Protective Services, and Faculty
Member, Saint Edwards University, Austin, Texas. Janess Sheets and Judith Henry are
Program Specialists, Texas Department of Family and Protective Services, Austin, Texas.
Victoria Jeffries is Doctoral Candidate, Simon Frazier University, Burnaby, Canada.
152 CHILD WELFARE • VOL. 87, #2
Findings indicate that even when controlling for risk and
poverty (as well as other relevant factors), race affects the
decision to provide services and to remove children. We
find that poverty is associated with higher risk scores. We also find
that the risk scores of African American families in cases that
are closed, those receiving Family Based Safety Services, and
those resulting in children being removed are lower than the
risk scores for Anglo families in the same groups. This suggests
that rather than racial bias in the assigning of the risk score
itself, disproportionality may be better explained by racial/ethnic
differences in the risk threshold workers use to decision to take
action on a case. In particular, the risk threshold for providing
services or removing a child is higher for Anglo Americans than
for African Americans.
The young caseworker stands at the doorway of the home. As
she looks at her surroundings, she sees a home in some disrepair
and a neighborhood that has seen better times. From her drive
down the block, she recalls cracked sidewalks and streets, broken
and boarded windows on homes, graffiti, and groups of African
American youth standing on corners. She needs to make several
decisions soon that will affect the safety of the child she is about to
see. The intake call sounded serious. Aside from deciding whether
or not to substantiate the case, she must determine whether the
risk of future harm is sufficient to provide services or regrettably,
if high enough, to place the child in foster care. How will she make
this decision? How will the poverty she has seen affect this judg-
ment? Will race play a role? Will she confuse them with risk?
It is important to know the answers to these questions because
if the judgment she makes is unduly influenced by these external
factors, we would want to know this so that steps can be taken to
improve the uniformity and consistency of caseworkers’ responses.
Knowing these factors would allow us to do two things. First, if
Address reprint requests to Donald Baumann, Texas Department of Family and Protec-
tive Services, Mail Code W157, P.O. Box 149030, Austin, TX 78714-9030. E-mail:
Rivaux et al. 153
we could determine that the risk instrument is racially biased, we
could repair it. Second, if the decision error is a misjudgment asso-
ciated with the way in which she thinks about risk, race, and
poverty, it is possible that we can inoculate her against this
through training or other strategies. An example of one type of
error that may be implicated here is the fundamental attribution
error, which refers to the tendency to underestimate the influence
of situational forces in the lives of those we are observing and to
overestimate the influence of personal factors, such as traits and
attitudes (Ross, 1977). In the example of our caseworker, it means
that poverty (a situational factor) is discounted in decision making
in favor of race (a personal factor).
How might we proceed in order to learn more about how risk,
race, and poverty may impact our caseworker’s decision to close,
open the case for services, or remove the child? Research in this area
has focused on identifying a racial difference in the rate at which
children in the Child Protective Services system are removed or
given Family Based Safety Services (FBSS).
These researchers have
found that African American children are more likely to be removed
from their homes and less likely to be given FBSS than Anglos
(Curtis, Dale, & Kendall, 1999; Garland, Hough, et al., 2000; Garland,
Landsverk, & Lau, 2003; Hill, 2006; Sedlak & Broadhurst, 1996;
Stoltzfus, 2005; Texas Health and Human Services Commission and
Texas Department of Family and Protective Services, 2006; U.S. Chil-
dren’s Bureau, 1997; U.S. Department of Health and Human Services
[USDHHS], 2005, 2006; Wulczyn, Barth, Yun, Jones-Harden, &
Studies have also examined factors that may explain dispro-
portionate representation of African Americans in more intrusive
intervention case decisions and findings from these studies are
mixed regarding the role of race. Examples of factors that have
been studied as possible explanations of the race/case decision as-
Acknowledgment: We wish to thank two anonymous reviewers who provided helpful
comments that improved this article.
1 In many parts of the United States, this is known as Family Preservation Services.
154 CHILD WELFARE • VOL. 87, #2
sociation include maltreatment reason, maltreatment history,
poverty, single parenthood, neighborhood crime level, and parental
employment. Some studies find no race effect when controlling for
these factors (Ards, Chung, & Myers, 1998; Harris, Tittle, & Poert-
ner, 2005; Katz, Hampton, Newberger, Bowles, & Synder, 1986;
Lindsay, 1994; Runyan, Gould, Trost, & Loda, 1981; Sedlak &
Broadhurst, 1996; Texas Health and Human Services Commission
and Department of Family and Protective Services, 2006; U.S. Chil-
dren’s Bureau, 1997) However, other studies find that race is a pre-
dictor of placement in and exit from foster care even when other
factors are taken into account (Chibnall et al., 2003; Hill, 2005;
Lu, Landsverk, Ellis-MacLeod, Newton, Ganger, & Johnson, 2004;
Needell, Brookhart, & Lee, 2003; Sedlack & Schultz, 2005; USD-
HHS, 2005; Wulczyn et al., 2005). One possible reason for this in-
consistency in the literature is that measures of risk are not always
included or adequately measured in some studies. Another possible
reason is that studies do not always separate the factors identify-
ing risk and factors affecting the decision to take action in a case.
In this study, we explicitly explored the relationship between race,
poverty, and risk for the decision to take action on a case (to provide
FBSS services or remove a child). To do so we employed a multi-
variate model that used these three factors along with others found
to be significant in prior studies. We also examined the role of risk
thresholds in explaining the findings.
Our first hypothesis was that, above and beyond risk and income,
race would predict whether further action would be taken on a
case. Specifically, we hypothesized that Anglo American families
would be more likely to have their cases closed while African
American families would be more likely to have their cases acted
upon. Our second hypothesis was that race would predict the type
2 We included these other variables as controls because they had a known association with our crite-
rion and could possible covary with our variables of interest.
Rivaux et al. 155
of CPS action taken, and specifically, that when risk levels, income,
and other factors are controlled for,
Anglo American families
would be more likely to receive FBSS than are African American
families, and African American children would be more likely to
be removed from their homes than are Anglo children. Factors that
we controlled for in these analyses included sociodemographic
variables, variables related to the initial report to the Department
of Family and Protective Services (DFPS), and variable interactions
that are explained in detail.
Three analyses were conducted on cases that reached the end of an
investigation. In the first, mean differences in risk scores were ex-
amined as a function of race and income. In the second and third,
we tested the previously stated hypotheses.
The overall sample was cases assigned to decisions to (1) take
no action/close a case, (2) provide FBSS, or (3) remove the child to
foster care or other substitute care. The sample included cases from
the Texas DFPS child welfare database from the period of September
1, 2003, through February 28, 2005 (n 123,621). Because referrals
to DFPS often involve multiple victims in one family, the investi-
gators used the family or “case” as the unit of analysis. Where fam-
ilies had more than one investigation in the time frame, each inves-
tigation was treated as a separate case.
First, ANOVA and Tukey post hoc tests were used to examine
the mean differences in risk scores by racial and income groups in
the full sample. Then, bivariate logistic regression models were used
to test the two hypotheses about factors that might differentially
predict taking action on a case and the FBSS versus removal deci-
sions. Logistic regression analysis allows for examination of poten-
tial predictors (e.g., race, risk, income) of a dichotomous outcome
(e.g., whether to close a case or act on it, whether to provide FBSS or
remove). So that inferential statistics could be more easily interpreted,
two subsamples were drawn consisting of only African American
and Anglo American cases. An approximately 25% subsample was
156 CHILD WELFARE • VOL. 87, #2
drawn from all cases to determine whether disproportionality exists
in the decision to take action on a case (n 26,846). To determine
whether disproportionality exists in decisions to provide FBSS or re-
move children, an approximately 25% subsample was drawn from
cases resulting in one of these decisions (n 4,027).
Potential explanatory variables for all tests of hypotheses were
identified through both literature review and availability. The main
variables included race/ethnicity, household income, and the case-
worker’s risk assessment score after investigating the report. The
risk assessment score is a composite score constructed by summing
seven risk Areas of Concern reported by caseworkers after they in-
vestigate a reported allegation of abuse/neglect. The seven Areas of
Concern are each scored separately on five point scales where one
is rated “not at all a concern” and five is rated “an extreme concern”
The areas are child vulnerability, caregiver capability, quality of
care, maltreatment pattern, home environment, social environment,
and response to intervention.
Other variables included socio -
demographic variables and variables related to the initial report to
DFPS. Sociodemographic variables included age group of the fam-
ily’s youngest child, number of children, number of alleged victims
and alleged perpetrators, parents’ marital status, gender, and whether
at least one parent was a teen parent. Report variables included
type of allegation, source of the report, and state region of report. In
addition, since disproportionality may be potentially explained
through interacting variables, interaction terms were created. The
interactions included in the models were race by income and low
income by whether the case was a neglect case.
All hypothesized explanatory variables were entered into ini-
tial logistic regression models and, following principles of parsi-
mony, any that did not contribute significantly to the model were
eliminated in the final model (p .01 in the likelihood ratio tests).
Variables found to have low explanatory power and excluded
from the final model were number of alleged victims and number
3 The scales have demonstrated both reliability and validity (Baumann et al., in press).
Rivaux et al. 157
of alleged perpetrators.
The total sample was 33.8% African American and 66.2% Anglo
American. Single parent households comprised 71.7% of the sam-
ple, families headed by mothers comprised only 40.4% of the sam-
ple, and families headed by a teen parent comprised 5.0%. Many
families were quite poor with 31.4% of families having a household
income of less than $10,150. Most families had more than one child
(68.5%), only one reported victim (62.8%), and only one perpetrator
(64.5%). The initial reports to DFPS came from legal/medical/or
CPS itself (34.2%), schools or day care centers (18.0%), the victim,
relatives, or friends (30.2%), and other sources (10.3%); anony-
mous reports accounted for 7.3% of all reports. The types of allega-
tions included neglect (34.9%), sexual abuse only (10.5%), physical
abuse only (24.7%), abandonment (1.0%), or multiple forms of
abuse (28.8%). Of the sample used for analysis (n 123,621), 9,635
(7.8%) cases resulted in FBSS, 6,352 (5.1%) resulted in removal to
foster care or some other type of substitute care, and 107,634
(87.1%) cases were closed after investigation.
Figure 1 displays the mean risk scores by a race and income for
the full sample. As indicated in the figure, there are different risk
assessment levels by both race and income within each decision
category. Specifically, (1) risk scores for FBSS are higher than
closed cases and risk scores for removal are higher than FBSS
cases, (2) the risk scores for lower income groups at all decision
levels were higher than those with higher incomes, and (3) risk
scores for Anglo Americans at all decision levels were higher than
risk scores for African Americans (all effects are significant beyond
The first finding (i.e., that risk scores are
4 We also ran a multiple regression equation order to determine the top predictors of risk. The order is
as follows with the beta weights in parentheses: (1) multiple abuse types (1.56), (2) law enforcement
reporter (.67) (3) race (.61), (4) physical abuse (.56), and (5) income (.53).
158 CHILD WELFARE • VOL. 87, #2
graduated from lowest to highest based on whether the case was
closed, resulted in FBSS, or resulted in removal) is what would be
expected. However, the other two findings suggest potential inter-
actions between risk assessment scores, race/ethnicity, and in-
come. Specifically, these findings suggest that those with lower
incomes were generally rated as being at higher risk and that, even
within comparable case decision categories, Anglo Americans had
higher risk than African Americans.
Table 1 presents results of the bivariate logistic regression analy-
sis testing the hypothesis that all else being equal, Anglo Americans
5 Logistic regression is used to understand predictors of dichotomous outcomes, in this case the deci-
sion to close a case versus the decision to act upon it (to offer FBSS services or remove one or more
children). The output of this analysis is an odds ratio. A number greater that one refers to a variable
being more likely to affect this outcome and a number less than one means that it is less likely.
Risk Scores by Race and Income
Mean Risk Score
Risk Scores by Race and Income
Rivaux et al. 159
Bivariate Logistic Regression Results for Closed Versus Taking Action
and FBSS Versus Removed
CLOSED VS. ACTION FBSS VS. REMOVAL
OR SIG. 95% CI OR SIG. 95% CI
LOWER UPPER LOWER UPPER
Risk Assessment Score
.001 1.363 1.392 1.253
.001 1.228 1.278
Less than $10,150 2.40
.001 1.898 3.035 0.99 0.980 0.62 1.60
.001 1.315 2.031 0.57 0.016 0.36 0.90
$20,550–40,549 1.166 0.175 0.934 1.457 0.76 0.254 0.47 1.22
$40,550 and greater*
African American 1.208 0.005 1.06 1.378 1.77
.001 1.36 2.30
Age of Oldest Child
Less than 1 year 3.143
.001 2.730 3.619 1.38 0.014 1.07 1.77
1–2 years 1.579
.001 1.371 1.818 0.91 0.462 0.69 1.18
3–5 years 1.160 0.039 1.008 1.335 0.81 0.147 0.62 1.08
13–16 years 0.793 0.007 0.669 0.939 2.64
.001 1.89 3.69
17 years 0.422 0.005 0.230 0.771 7.43 0.003 1.94 28.45
Number of Children
Multiple children 1.096 0.101 0.982 1.222 0.58
.001 0.48 0.70
Only one child*
Family Gender Composition
Mixed 1.177 .013 1.035 1.338 0.98 0.885 0.78 1.24
All female 1.089 0.119 0.978 1.211 0.86 0.137 0.71 1.05
Parents’ Marital Status
Not married 0.849 0.004 0.760 0.948 1.34 0.005 1.10 1.64
Age of Parent
Teen parent 1.110 .268 .923 1.335 0.67 0.008 0.50 0.90
Not teen parent*
High Plains/Upper 1.158 0.203 0.924 1.45 0.95 0.779 0.66 1.36
Northwest/Upper 0.815 0.039 0.671 .990 0.67 0.025 0.47 0.95
Upper East/ 1.088 0.284 0.933 1.268 0.86 0.284 0.65 1.13
Gulf Coast, 2.478
.001 2.177 2.821 0.97 0.783 0.77 1.22
160 CHILD WELFARE • VOL. 87, #2
TABLE 1 cont.
CLOSED VS. ACTION FBSS VS. REMOVAL
OR SIG. 95% CI OR SIG. 95% CI
LOWER UPPER LOWER UPPER
Central 1.015 0.849 0.868 1.187 1.16 0.288 0.88 1.52
Upper South 2.483 .001 2.070 2.977 0.46
.001 0.33 0.65
Lower South 2.780 .001 2.153 3.590 0.48 0.004 0.29 0.79
Dallas Ft. Worth
Legal/Medical/FPS 1.917 .001 1.709 2.149 1.99
.001 1.60 2.48
School/Daycare 1.332 .001 1.137 1.561 1.25 0.175 0.91 1.72
Anonymous 0.881 0.260 0.707 1.098 1.03 0.900 0.67 1.57
Other 1.081 0.392 0.904 1.293 2.04
.001 1.48 2.80
Sexual abuse only 0.923 0.511 0.742 1.161 0.48 0.010 0.28 0.84
.001 3.924 8.123 14.70
.001 7.06 30.62
Multiple types of 1.391
.001 1.205 1.607 1.01 0.945 0.77 1.33
Physical abuse only 1.074 0.377 0.917 1.259 0.68 0.018 0.50 0.94
Neglect Income Interaction
Other 0.905 0.309 0.746 1.097 0.97 0.858 0.69 1.37
Both neglect and
Race Income Interaction
Low-income 0.916 0.365 0.757 1.108 0.80 0.187 0.57 1.12
High-income ** ** ** ** ** ** ** **
Low-income ** ** ** ** ** ** ** **
Significant ORs are in bold.
*Indicates reference category
**Indicates insufficient cell size for analysis.
Rivaux et al. 161
would be more likely to have their cases closed than would African
The table includes the odds ratios and 95% confidence
intervals estimating the relationship between predictors and case
service decisions. In logistic regression, odds ratios estimate the
probability of a given outcome for different groups. If, for example,
racial/ethnic groups are being compared on a given outcome and
an odds ratio greater than one is obtained for a particular
racial/ethnic group, this would mean that the group is more likely
to have that outcome than are other groups. Conversely, if an odds
ratio is less than one for that racial/ethnic group, this would mean
that the outcome is less likely for that group relative to other groups.
The final model demonstrated acceptable model fit (
8411, p < .001). Race did contribute to the decision to take action on
a case; specifically, when compared to Anglo Americans, African
Americans were 20% more likely to have their case acted upon.
Our theoretical variables, risk and income, were also significant
predictors. Other variables that contributed significantly to the
service decision were child’s age, parents’ marital status, report
source, allegation type, and the family living in certain regions of
the state. Since this was a subsample, split sample validation was
run to validate the results. The validation also demonstrated ac-
ceptable model fit (
(32) 8251, p .001), but race was not sig-
nificant in this model. However, when analyzing the full data set
(32) 33,265, p .001), the race variable was sig-
nificant in that African Americans were 12% more likely to have
their cases acted upon than Anglo Americans.
Table 1 also presents results of the bivariate logistic regression
analysis testing the hypothesis that all else equal, when actions are
taken, Anglo Americans would be more likely to be assigned to
FBSS while African Americans would be more likely to be re-
moved. The table includes the odds ratios and 95% confidence in-
tervals estimating the relationship between predictors and case
service decisions. The final model demonstrated acceptable model
(32) 1223.0, p .001) and resulted in a significant effect of
race on the decision about what actions would be taken; specifi-
162 CHILD WELFARE • VOL. 87, #2
cally, when compared to Anglo Americans, African Americans
were 77.0% more likely to be removed. Once again, our theoretical
variables, risk and income, were significant. Other variables that
contributed significantly were child age, whether the parents were
married, level of income, number of children in the family, whether
the parents were teens, report source, allegation type, and state re-
gion of the report. Since this was a subsample, split sample valida-
tion was run to validate the results. The validation also dem -
onstrated acceptable model fit (
(32) 1295.92, p .001) and
resulted in a similar significant effect of race on these decisions, that
is, that African Americans were 87.3% more likely to be removed
than were Anglo Americans. Additionally, with the exception of
parents’ marital status no longer showing significance, there was a
similar pattern of significant covariates as in the initial model.
Other previous research suggests that racial differences in case
decisions may be explained by risk assessment instruments that are
racially or economically biased, rather than by possible bias on the
part of caseworkers (e.g., Brissett-Chapman, 1997). If a racial bias
were present, then one would expect the risk scores of African
American families would be higher than Anglo families, since they
are more likely to be removed. This was not the case: African Amer-
ican families actually were found to have lower risk scores when re-
moved than Anglo American families (see Figure 1), though risk and
poverty were found to be related.
The bivariate logistic analyses suggest that, all else being equal,
race, risk, and income predict the services decision and the removal
decision. Findings on risk did indicate that families whose children
were removed were rated at higher risk than families who were
provided services, who were, in turn, rated at higher risk than fam-
ilies whose cases were closed. Results further showed that families
with low income had higher risk scores than families with higher
income, and that within decision categories, African American fam-
ilies were rated at lower risk than Anglo families. Finally, these effects
e.g. Risk or ‘Level of concern’
The Case Factors
History of Decision
maker (Race or
If the Assessment is ABOVE the Threshold, then ACTION is taken.
If the Assessment is BELOW the Threshold, then NO ACTION is taken.
Rivaux et al. 163
could not be attributed to a cultural bias in risk assessment, yet how
poverty and risk are related remains to be seen.
How might these effects on risk operate in relation to the deci-
sions we are describing? Dalgleish (2003, 2006) argues from a sig-
nal detection framework (Tanner & Swets, 1954) that individuals’
assessments of risk can be similar, but their decision thresholds
may differ. He further argues that the factors that influence the as-
sessment are those associated with information from the current
situation, that is, the case factors; those factors influencing the
decision threshold are those from the decision makers’ history or
experience. We have displayed this relationship in Figure 2.
A General Model for Assessing the Situation and Deciding What To Do About It—
6 We have modified the model slightly to both display the factors in this discussion in parentheses.
164 CHILD WELFARE • VOL. 87, #2
As shown in Figure 2, we view income as a case factor that in-
fluences risk assessment, but not a factor that influences the
threshold for the decision. We contend (as would Dalgleish) that
the threshold itself is affected by variables associated with the de-
cision maker such as their views on race. As indicated in our find-
ings, this would mean that even though African Americans are as-
sessed as having lower risk scores, they are more likely to have a
different threshold for removals and service provision than are An-
glos. We take the model further, however, in suggesting that other
forces can operate to shift these thresholds. This is consistent with
a decision-making ecology framework (Baumann, Fluke, & Kern,
1997), whereby factors that are associated with the organization or
the community might alter one’s threshold. We list one such exam-
ple on the right side of Figure 2: service availability. That is, the
lack of available services in African American communities could
possibly be responsible for the difference in decision-making that
we have shown. Specifically, the families’ problems were sufficient
to warrant taking action and if there were no available services, a
removal may have been the only choice the worker had (i.e., the
threshold was lowered).
In introducing this paper, we noted that some studies find ef-
fects for race and others do not, sometimes showing an effect for
poverty instead. The present findings and the model we provide
suggests why this inconsistency may exist. Poverty may serve as
an indicator of risk. However, when race is included in the model,
race changes the decision threshold.
Limitations and Future Research
The tests that we provide of this model are only a beginning. We
suggest three avenues in future studies that would help illumi-
nate the model we offer and offset the limitations in the present
model regarding determining causality. The first would involve
analyses that control for hierarchical effects that may exist in this
study due to assessing multiple cases from the same family. The
Rivaux et al. 165
second would involve using structural equation modeling
whereby caseworker variables (e.g., cultural sensitivity) would
be associated with what is referred to as a disparity index built
from case level data (Fluke, Parry, & Baumann, 2006). This index
specifies the degree to which a particular decision is based upon
race (all else being equal). Community (e.g., service availability)
and organizational (e.g., workload) factors could be associated
with this same index and the relative contribution of all of these
forces on the decision could be tested. This forms the basis of the
decision-making ecology and allows one to empirically under-
stand what factors are influencing these decisions and design in-
The third avenue that is likely to bear fruit is an experimental
examination of how poverty, risk, and race affect decision thresh-
olds more directly. Clearly, poverty and risk are related. Neverthe-
less, poverty is situational and perhaps should be a part of how we
understand risk. The difference is that our response to poverty also
needs to be situational. As we indicated in the introduction to this
paper, poverty, risk, and race may be related due to the fundamen-
tal attribution error through which decisions may be made based
on an underestimation of situational forces such as poverty. If
more on this error in thinking could be uncovered, perhaps train-
ing or other strategies could be developed that would inoculate
decision makers against it.
Findings from this study indicate that even when controlling for
risk, poverty, and other relevant factors, race affects the decisions to
provide services and to remove. Poverty was associated with
higher risk scores and Anglo Americans were more likely to have
higher risk scores within case decision categories. These findings
suggest that, rather than racial bias in the assigning of the risk score
itself, disproportionality may be better explained by racial/ethnic
differences in the risk threshold workers use to make case deci-
166 CHILD WELFARE • VOL. 87, #2
sions. In particular, the risk threshold for more intrusive case deci-
sions is higher for Anglo Americans than for African Americans.
Ards, S. D., Chung, C., & Myers, S. (1998). The effects of sample selection bias on racial dif-
ferences in child abuse reporting. Child Abuse and Neglect, 22(2), 103– 115.
Baumann, D. J., Fluke J., & Kern, H. (1997). Workers improvements to the structured decision and
outcome model (WISDOM): The child welfare decision enhancement project. Austin: The Texas
Department of Family and Protective Services.
Brissett-Chapman, S. (1997). Child protection risk assessment and African American chil-
dren. Child Welfare, 76(1), 45– 63.
Chibnall, S., Dutch, N. M., Jones-Harden, B., Brown, A., Gourdine, R., Boone, A., & Snyder, S.
(2003). Children of color in the child welfare system: Perspectives from the child welfare commu-
nity. Washington, DC: U.S. Department of Health and Human Services, Government
Curtis, P., Dale, G., & Kendall, J. (Eds.) (1999). The foster care crisis: Translating research into
policy and practice. Lincoln: University of Nebraska Press.
Dalgleish, L. I. (2003). Risk, needs and consequences. In M. C. Calder (Ed.), Assessments in
child care: A comprehensive guide to frameworks and their use (pp. 86– 99). Dorset, UK: Rus-
sell House Publishing.
Dalgleish, L. I. (2006, September). Testing for the effects of decision bias on overrepresenta-
tion: Applying the GADM model. Paper presented at the 16th International Congress on
Child Abuse and Neglect.
Fluke, J., Parry, C., & Baumann, D. J. (2006, September). Multi-national research applications
of the decision making ecology (DME) model and general assessment and decision-mak-
ing models to the issue of racial and ethnic over-representation of protective services:
Modeling decision factors in the U.S. Paper presented at the 16th International Congress on
Child Abuse and Neglect.
Garland, A., Hough, R., Landsverk, J., McCabe, K., Yeh, M., Ganger, W., & Reynolds, B.
(2000). Racial/ethnic differences in mental health care utilization among children in fos-
ter care. Children’s Services: Social Policy, Research and Practice, 3, 133– 146.
Garland, A., Landsverk, J., & Lau, A. (2003). Racial/ethnic disparities in mental health serv-
ice use among children in foster care. Children and Youth Services Review, 25(5/6), 491– 507.
Harris, G., Tittle, G., & Poertner, J. (2005). Factors that predict the decision to place a child
Rivaux et al. 167
into substitute care. In D. Derezotes, J. Poertner, & M. Testa (Eds.), Race matters in child
welfare: The overrepresentation of African American children in the system (pp. 163– 172).
Washington, DC: Child Welfare League of America.
Hill, R. B. (2005). The role of race in foster care placement. In D. Derezotes, J. Poertner, &
M. Testa (Eds.), Race matters in child welfare: The overrepresentation of African American chil-
dren in the system (pp. 187– 200). Washington, DC: Child Welfare League of America.
Hill, R. B. (2006). Synthesis of research on disporportionality in child welfare: An update. Retrieved
February 11, 2007, from http://www.cssp.org/whatsnew.html#recent
Katz, M., Hampton, R., Newberger, E., Bowles, R., & Synder, J. (1986). Returning children
home: Clinical decision making in cases of child abuse and neglect. American Journal of
Orthopsychiatry, 56(2), 253– 262.
Lindsay, D. (1994). Factors affecting the foster care placement of decision: An analysis of
national survey data. American Journal of Orthopsychiatry, 6(2), 272– 281.
Lu, Y. E., Landsverk, J., Ellis-MacLeod, E., Newton, R., Ganger, W., & Johnson, I. (2004).
Race, ethnicity, and case outcomes in child protective services. Children and Youth Serv-
ices Review, 26(5), 447– 461.
Needell, B., Brookhart, A., & Lee, S. (2003). Black children and foster care placement in Cal-
ifornia. Children and Youth Services Review, 25(5/6), 393– 408.
Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribu-
tion process. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 10,
p. 88). New York: Academic Press.
Runyan, D., Gould, C., Trost, D., & Loda, F. (1981). Determinants of foster care placement
for the maltreated child. American Journal of Public Health, 71(7), 706– 711.
Sedlak, A. J., & Broadhurst, D. D. (1996). Executive summary of the third national Incidence
Study of Child Abuse and Neglect [Electronic version]. Washington, DC: U.S. Department
of Health and Human Services.
Sedlak, A., & Schultz, D. (2005). Racial differences in child protective services investigation
of abused and neglected children. In D. Derezotes, J. Poertner, & M. Testa (Eds.), Race
matters in child welfare: The overrepresentation of African American children in the system
(pp. 97– 118). Washington, DC: Child Welfare League of America.
Stoltzfus, E. (2005). Race/ethnicity and child welfare. Washington, DC: Congressional Research
Tanner, W. P., & Swets, J. A. (1954). A decision-making theory of visual detection. Psycholog-
ical Review, 61, 401– 409.
168 CHILD WELFARE • VOL. 87, #2
Texas Health and Human Services Commission and Texas Department of Family and Pro-
tective Services. (2006). Disproportionality in child protective services: Statewide reform effort
begins with examination of the problem. Austin, TX: Author.
U.S. Children’s Bureau. (1997). National study of protective, preventive, and reunification services
delivered to children and their families. Washington, DC: U.S. Department of Health and
U.S. Department of Health and Human Services, Administration for Children, Youth, and
Families. (2005). AFCARS report #10. Washington, DC: U.S. Government Printing Office.
U.S. Department of Health and Human Services, Administration on Children, Youth, and
Families. (2006). Child maltreatment 2004. Washington, DC: U.S. Government Printing
Wulczyn, F., Barth, R., Yuan, Y., Jones-Harden, B., & Landsverk, J. (2005). Beyond common
sense: Child welfare, child well-being and the evidence for policy reform. New Brunswick, NJ: