Children and Youth Services Review 41 (2014) 95–105

Contents lists available at ScienceDirect

Children and Youth Services Review

journal homepage: www.elsevier.com/locate/childyouth

Neighborhood racial & ethnic diversity as a predictor of child welfare

system involvement
Sacha Klein a, Darcey H. Merritt b,⁎
a
Michigan State University, School of Social Work, United States
b
New York University, Silver School of Social Work, United States

a r t i c l e i n f o a b s t r a c t

Article history: Using U.S. Census and child maltreatment report data for 2052 Census tracts in Los Angeles County, California,
Received 6 May 2013 this study uses spatial regression techniques to explore the relationship between neighborhood social disorgani-
Received in revised form 10 March 2014 zation and maltreatment referral rates for Black, Hispanic and White children. Particular attention is paid to the
Accepted 13 March 2014
racial–ethnic diversity (or ‘heterogeneity’) of neighborhood residents as a risk factor for child welfare system
Available online 21 March 2014
involvement, as social disorganization theory suggests that cultural differences and racism may decrease neigh-
Keywords:
bors' social cohesion and capacity to enforce norms regarding acceptable parenting and this may, in turn, increase
Child maltreatment neighborhood rates of child maltreatment. Results from this study indicate that racial–ethnic diversity is a risk
Diversity factor for child welfare involvement for all three groups of children studied, even after controlling for other indi-
Ecological models cators of social disorganization. Black, Hispanic and White children living in diverse neighborhoods are signifi-
Neighborhoods cantly more likely to be reported to Child Protective Services than children of the same race/ethnicity living in
Race more homogeneous neighborhoods. However, the relationships between child welfare system involvement
Spatial analysis and the other indicators of social disorganization measured, specifically impoverishment, immigrant concentra-
tion child care burden, residential instability, and housing stress, varied considerably between Black, Hispanic
and White children. For Black children, only housing stress predicted child maltreatment referral rates; whereas,
neighborhood impoverishment, residential instability, and child care burden also predicted higher child mal-
treatment referral rates for Hispanic and White children. Immigrant concentration was unrelated to maltreat-
ment referral rates for Black and Hispanic children, and predicted lower maltreatment referral rates for White
children. Taken together, these findings suggest that racial–ethnic diversity may be one of the more reliable
neighborhood-level demographic indicators of child welfare risk across different racial/ethnic groups of children.
However, many of the other neighborhood characteristics that influence child maltreatment referrals differ for
Black, Hispanic and White children. Consequently, neighborhood-based family support initiatives should avoid
a one-size-fits-all approach to child abuse prevention and strategically consider the racial/ethnic make-up of
targeted communities.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction illuminate the relationship between neighborhood structure and child

A growing body of research links several neighborhood structural
factors to child maltreatment rates (see Coulton, Crampton, Irwin,
Spilsbury, & Korbin, 2007; Freisthler, Merritt, & LaScala, 2006), suggest-
ing that place-based interventions may be a promising avenue for
reducing child abuse and neglect. There has also been a great deal of
research contributing to the discussion regarding racial and ethnic
disproportionality among children exposed to the U.S. child welfare sys-
tem (Fluke, Jones-Harden, Jenkins, & Ruehrdanz, 2010; Fluke, Yuan,
Hedderson, & Curtis, 2003; Hill, 2006; Kim, Chenot, & Ji, 2011; Stoltzfus,
2005; Wulczyn & Lery, 2007). Yet, only a few studies have attempted to
⁎ Corresponding author at: 1 Washington Square North, 316, New York, NY, 10003,
United States.
welfare system involvement for children of different races/ethnicities
(Freisthler, Bruce, & Needell, 2007; Kohl, Jonson-Reid, & Drake, 2009;
Korbin, Coulton, Chard, Platt-Houston, & Su, 1998). The current study
expands on this research by examining the relationship between child
maltreatment referral rates and neighborhood racial–ethnic diversity
(or ‘heterogeneity’). Separate models are presented for Black, Hispanic
and White children living in Los Angeles County, California.
1.1. Racial/ethnic disproportionality in U.S. child welfare system
involvement
Although national data on the race and ethnicity of children referred
to child protective services (CPS) in the United States is not readily
available, recent statistics on substantiated referrals demonstrate that
among the 93.3% of substantiated cases for which victim race is

http://dx.doi.org/10.1016/j.childyouth.2014.03.009
0190-7409/© 2014 Elsevier Ltd. All rights reserved.
96 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
specified, 93.7% are White, African American or Hispanic, while 4.1% are
classed as being of multiple race, and only 2.3% are identified as belong-
ing to other racial groups (e.g. American Indian, Asian, Pacific Islander).
The relative risk of maltreatment for each of these racial/ethnic groups
varies considerably. In Federal Fiscal Year 2011, Blacks were substantial-
ly over-represented among confirmed child maltreatment victims,
Whites were under-represented, and Hispanics made up almost the
same percentage of maltreatment victims as children in the general
population (United States Department of Health et al. [USDHHS],
2011). Similar to national trends, child welfare system involvement in
Los Angeles County, where the current study was conducted, is also
characterized by racial and ethnic disproportionality. The most recent
data available demonstrate that only Hispanics and Native Americans
are referred to child protective services (CPS) at roughly the same rate
as their presence in the county's general population (e.g. Hispanics
make up 61% of Los Angeles County's child population and also account
for 61% of the referrals to CPS). Asian Americans and Whites are under-
represented, while Black children are over-represented. Blacks account
for nine percent of the child population in Los Angeles and 19% of the
child maltreatment referrals (Needell, Brookhart, & Lee, 2003).
There has been much debate regarding the underlying causes of
racial/ethnic disproportionality in the child welfare system, with two
general schools of thought emerging. Drake, Jolley, Lanier, Fluke, and
Barth (2011) describe these as the “Bias Model” and the “Risk Model”.
The former attributes the over-representation of Blacks in the child wel-
fare system (and in some communities other minority groups) to racial
bias on the part of those who report and investigate maltreatment.
Inherent to the “Bias Model” is the assumption that minorities do not
actually mistreat their children more, or at least not to the extent sug-
gested by their disproportionate level of CPS contact. Rather, their
over-representation is understood to be the product of excessive scruti-
ny by community members and professionals who are prone to pathol-
ogizing and labeling questionable parenting by minorities as ‘abuse’ or
‘neglect’ due to racist presuppositions. The “Risk Model”, on the other
hand, contends that over-represented minorities have more child wel-
fare system contact because they do in fact maltreat their children
more often than members of other groups. According to this model,
over-represented racial/ethic groups engage in higher rates of child
maltreatment because they are, on average, exposed to more personal
and community-level risk factors, such as poverty and unemployment,
and tend to have less access to services and supports that can buffer
them against the parenting stress associated with exposure to these
risks. While research exists to support both theories, evidence
supporting the “Risk Model” is stronger (see Drake et al., 2011).
1.2. The impact of neighborhood
Largely missing from this theoretical debate is a discussion of how
neighborhoods, particularly the racial/ethnic structure of neighbor-
hoods, contribute to disproportionate child welfare system involve-
ment. Scholars have noted myriad neighborhood characteristics that
are related to child well-being, the potential for maltreatment, and/or
child welfare system involvement. Poverty rates, concentrated afflu-
ence, unemployment, percentage of female-headed households, com-
munity violence, urbanicity/rurality, residential turnover, population
density, residential crowding, vacant housing, immigrant concentration,
spatial relation to other communities, adequacy of informal resources
for child supervision (often termed ‘child care burden’), accessibility of
formal child care and early education services, and the presence of
local alcohol outlets are all implicated in neighborhood or community-
level studies of child abuse and neglect (Coulton & Korbin, 1995;
Coulton, Korbin, & Su, 1999; Coulton et al., 2007; Drake, Lee, &
Jonson-Reid, 2009; Drake & Pandy, 1996; Freisthler, Bruce, & Needell,
2007; Freisthler, Gruenewald, Remer, Lery, & Needell, 2007; Freisthler,
Gruenewald, Ring, & LaScala, 2008; Freisthler, Needell, & Gruenewald,
2005; Freisthler et al., 2006; Gillham et al., 1998; Kim et al., 2011;
Klein, 2011; Korbin et al., 1998; Merritt, 2009; Spearly & Lauderdale,
1983; Zuravin, 1986). Research also demonstrates that families of color
are more likely than White families to reside in distressed neighbor-
hoods characterized by impoverishment, crime and other community
characteristics associated with child maltreatment (Drake & Rank,
2009; Massey, 2004).
1.2.1. Social disorganization
This body of research is underpinned by social disorganization theo-
ry (Sampson, 2001; Shaw & McKay, 1969; Wilson, 1987, 1996), which
posits that rapid structural changes in the United States since the
1970s have eroded the levels of community organization in many
urban neighborhoods precipitating the clustering of a host of social
problems, including child maltreatment, in these areas. According to
this theory, neighborhoods that are socially disorganized lack a struc-
ture to help maintain social controls that allow community members
to realize shared goals, like child safety. Mutual trust, shared norms
and expectations for resident behavior, the availability of informal and
formal support services, and a sense of “collective efficacy” are thought
to equip residents to effectively exert social control regarding undesir-
able behavior in their community (Sampson, 2001; Shaw & McKay,
1969). Consequently, if community members are in disagreement
regarding what constitutes acceptable standards of behavior, are
mistrusting and socially isolated from their neighbors, or are unable to
realize acceptable standards of behavior, the resultant social disorgani-
zation may precipitate elevated incidences of child maltreatment and
child welfare system contact.
Social disorganization researchers often rely on demographic mea-
sures to serve as proxies for core theoretical constructs. Neighborhood
poverty and unemployment rates are frequently used indicators of dis-
organization because it is theorized that residents of poor communities
are too busy addressing their survival needs to regularly socialize with
neighbors and actively participate in civic life (Coulton et al., 1999;
Deccio, Horner, & Wilson, 1994; Freisthler, 2004; Freisthler,
Gruenewald, Remer, Lery, & Needell, 2007; Freisthler, Midanik, &
Gruenewald, 2004; Klein, 2011; Krishnan & Morrison, 1995; Merritt,
2009; Molner, Buka, Brennan, Holton, & Earls, 2003). Similarly, commu-
nities with high levels of ‘child care burden’ are presumed more disorga-
nized because residents are too overwhelmed with child supervision
responsibilities to build relationships with their neighbors (Coulton
et al., 2007; Greenman, Bodovski, & Reed, 2011; Guterman, Lee,
Taylor, & Rathouz, 2010). Child care burden is often represented by
the ratio of children to adults and the percent of elderly people living
in a neighborhood, as adult residents' ability to supervise children
may compete with their role as caregivers for aging family members.
Another frequently used demographic indicator of social disorganiza-
tion is population turnover rates. Social disorganization theorists reason
that residential churning makes it hard to get to know one's neighbors
and build relationships characterized by mutual trust, which tends to
take time (Freisthler, Gruenewald, Remer, Lery, & Needell, 2007;
Kasarda & Janowitz, 1974; Osgood & Chambers, 2000; Sampson, 1988;
Sampson & Groves, 1989). Likewise, the concentration of immigrants
in a neighborhood and racial–ethnic heterogeneity (diversity) are
sometimes used to signify social disorganization as culture and lan-
guage differences are expected to disrupt the development of shared
norms between neighbors regarding appropriate behavior, while
racism and anti-immigrant sentiments may impede the development
of socially cohesive relationships (Sampson & Groves, 1989; Shaw &
McKay, 1972; Thomas, 2011). It should be noted, however, that empir-
ical evidence that immigrant concentration contributes to neighbor-
hood disorganization and associated outcomes for residents, including
child maltreatment, is weak (Freisthler, Bruce, & Needell, 2007; Lee,
Martinez, & Rosenfeld, 2001; Martinez, 2002; Morenoff, Sampson, &
Raudenbush, 2001; Nielsen, Lee, & Martinez, 2005; Sampson &
Raudenbush, 1999).
 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
1.2.2. Racial/ethnic neighborhood structure and child maltreatment
A handful of neighborhood child maltreatment studies drawing on
social disorganization theory examine the racial/ethnic composition of
communities as a potential predictor of child abuse and neglect, al-
though they do not always clearly articulate how or why the racial/ethnic
structure of a neighborhood is expected to influence neighbors' capacity
to organize against child abuse. Results from these inquiries are some-
what mixed, but generally suggest that child maltreatment and child
welfare system involvement are more common in communities
inhabited by Blacks and Hispanics. The percentage of Blacks living in a
neighborhood has been associated with higher rates of reported child
maltreatment, substantiated maltreatment, and foster care placement
(Freisthler, Gruenewald, Remer, Lery, & Needell, 2007; Zuravin, 1986).
The concentration of Hispanic residents has also been associated with
higher neighborhood rates of substantiated child maltreatment, partic-
ularly physical abuse, as well as foster care placement, but not initial
child welfare system contact (reporting rates) (Freisthler, 2004;
Freisthler, Gruenewald, Remer, Lery, & Needell, 2007; Freisthler et al.,
2005). In contrast, the concentration of White residents in a community
appears unrelated to neighborhood child maltreatment rates (Drake &
Pandy, 1996; Freisthler, Gruenewald, Remer, Lery, & Needell, 2007).
Only a few studies consider the possibility that neighborhood charac-
teristics may not influence child maltreatment risk uniformly for all chil-
dren and thus examine the differential relationship between community
structure and child welfare outcomes by children's race/ethnicity. Korbin
et al. (1998) tackled the question of differential neighborhood effects by
comparing predominantly White versus predominantly Black Census
tracts in Ohio. They found that impoverishment predicted rates of
maltreatment referrals in both types of communities, and an interaction
between neighborhood impoverishment and residential instability
predicted lower rates of reported child maltreatment. However, the
relationship between impoverishment and child maltreatment referrals
was much weaker in Black neighborhoods, while child care burden and
residential instability were related to maltreatment solely in the
White neighborhoods. Like Korbin et al. (1998), Freisthler, Bruce, and
Needell's (2007) analysis of 940 Census tracts in three California counties
found that poverty rates predicted substantiated child maltreatment for
all racial/ethnic groups studied: Black, Hispanic and White; while, immi-
grant concentration was not related to substantiated maltreatment for
any of these groups. Other indicators of social disorganization varied in
their relationship to child abuse/neglect outcomes across the different
races/ethnicities. The density of alcohol outlets (e.g. bars, liquor stores)
in the neighborhood, smaller population size, residential stability, and
low concentration of Black residents predicted substantiated maltreat-
ment of Black children. The percentage of female-headed families, unem-
ployment rate, and low concentration of Black residents predicted
substantiated maltreatment of Hispanic children. Child care burden
and higher concentration of Hispanic residents predicted substantiated
maltreatment of White children.
More recently, Drake et al. (2009) conducted an extensive analysis
of predictors of child maltreatment reports for Blacks and Whites living
in Missouri that focused on neighborhood poverty and racial composi-
tion. They found that poverty predicted higher rates of reporting for
both races. However, they also detected evidence of ‘differential sensi-
tivity’ in that White children were at higher risk than Black children of
being referred to child protection authorities when living in poor com-
munities, while Black children were at higher risk of being referred
when living in more affluent (less poor) communities. Drake et al.
(2009) also found that in Missouri “being out of place” (i.e. a minority
in your Census tract) was associated with higher maltreatment rates
for Whites and for Blacks in tracts with less than 15% Black children in
poverty, although these types of tracts were relatively rare, and so
they concluded that it was a contributing factor but a minor one in
explaining disproportionality.
Our study seeks to extend the understanding of how the communities
in which families live may contribute to racial/ethnic disproportionality
97
in child welfare system involvement by paying particular attention to
the influence of racial–ethnic diversity or ‘heterogeneity’, which is a di-
mension of neighborhood racial/ethnic structure that social disorganiza-
tion researchers have repeatedly documented as a risk factor for other
negative neighborhood outcomes such as elevated rates of juvenile delin-
quency and adult crime (Osgood & Chambers, 2000; Sampson & Groves,
1989; Sunghoon & Choo, 2008), but which has been heretofore under-
explored in relationship to child maltreatment. Instead, ecological studies
of child welfare outcomes have tended to examine the concentration of
specific racial/ethnic groups in neighborhoods (e.g. the percentage of
residents who are Black) as a risk factor rather than the overall degree
of diversity irrespective of the particular racial/ethnic groups (Ernst,
2001; Freisthler, Bruce, & Needell, 2007; Freisthler, Gruenewald, Remer,
Lery, & Needell, 2007; Freisthler et al., 2004; Freisthler et al., 2005).
1.3. Racial–ethnic diversity (a.k.a. ‘heterogeneity’)
We expect racial–ethnic heterogeneity will predict higher rates of
child maltreatment referrals for Black, Hispanic, and White children. As
previously noted, racial–ethnic heterogeneity (diversity) is a key con-
struct in social disorganization theory and is typically characterized as
the mix of different race/ethnic groups co-existing in defined geograph-
ical areas. Residence in a racially/ethnically diverse neighborhood is con-
sidered a risk factor for negative community outcomes because it may
decrease residents' social cohesion and capacity to enforce prosocial
norms (Osgood & Chambers, 2000; Sampson & Groves, 1989), including
norms regarding acceptable parenting. People tend to distrust their
neighbors when they are different from them, which means they are
less likely to rely on them for support/help and are less likely to agree
about and band together to address threats to the community (e.g. per-
ceived poor behavior and parenting). As Robert Putnam (2007, p. 137)
notes, diversity “seems to bring out the turtle in all of us.” In racially/
ethnically heterogeneous neighborhoods where the population lacks a
shared identity, residents may not be comfortable with diversity and
may ‘hunker down,’ trusting in others and participating in civic life less
than those living in more homogeneous areas (Kissane, 2010). While
multiple child maltreatment studies measure the relationship between
the concentration of specific racial/ethnic groups and child maltreat-
ment, we are aware of only one other study (Klein, 2011) that uses this
more theoretically grounded measure of racial–ethnic diversity to pre-
dict maltreatment. This study found that racial–ethnic heterogeneity
predicted higher neighborhood rates of maltreatment for children
under the age of six, but it did not explore the relationship between
diversity and maltreatment for older children or racial/ethnic subgroups.
1.4. Current Study
Our study contributes to the literature on community-level contrib-
utors to racial/ethnic disproportionality in child welfare system involve-
ment by examining the relationship between racial–ethnic diversity
within neighborhoods, several other indicators of social disorganization,
and maltreatment reporting rates separately for Black, Hispanic and
White children under the age of eighteen. In addition, we address a
common limitation of community-level analyses of child maltreatment
by considering spatial effects. Most child maltreatment studies fail to
account for the spatial relationship between neighborhood charac-
teristics (Coulton et al., 2007), which can distort parameter esti-
mates, so our study presents a spatial effects model at the Census
tract level in order to arrive at better estimates of the relationship
between neighborhood structure and maltreatment for children of dif-
ferent races/ethnicities.
Based on previous empirical research and the theoretical literature
summarized above, the current study hypothesizes that racially/
ethnically diverse neighborhoods will have higher child maltreatment
referral rates for each of the racial/ethnic groups studied. We did not
enter into this analysis with any preconceived ideas about whether
98 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
this construct would explain variation in maltreatment referral rates or
if/how explanatory power might vary by child's race or ethnicity.
2. Methods
2.1. Sample and procedures
This research relies on a cross-sectional ecological analysis of the
spatial relationship between neighborhood structure and maltreatment
referral rates for Black, Hispanic and White children in neighborhoods,
defined by the 2052 contiguous Census tracts that compose Los Angeles
County, California. Los Angeles County is comprised of 2054 census
tracts, however the two tracts that make up Catalina Island were omit-
ted because of their spatial discontinuity with “mainland” Los Angeles.
Including these tracts would have potentially biased analytic results
and hindered a meaningful assessment of, and adjustment for spatial
effects. Given that the unit of analysis is neighborhoods, this study is
poised to contribute information about the relationship between
community-level characteristics and child maltreatment referral rates
for three racial/ethnic groups. We caution against interpreting our find-
ings found at this macro (broader) level to be also inferred to exist at a
more micro (i.e., individual or family) levels, thus committing an “eco-
logical fallacy” (Robinson, 1950).
Los Angeles encompasses 4752 square miles and is the most popu-
lous county in the United States. According to 2005–2009 American
Community Survey, which includes estimates that cover a time period
that roughly corresponds with data collection for the current study,
Los Angeles County had almost 10 million (9,785,295) people living in
it, 26.0% of whom were children. The residents were 47.3% Hispanic
(any race), and 28.9% were non-Hispanic White, 12.9% Asian, 8.6%
African American, 0.2% Native American, 0.2% Pacific Islander or Native
Hawaiian, 0.3% some other race and 1.6% multiracial. More than half
(56.1%) spoke a language other than English at home, and more than
a third (35.4%) were foreign born. The median family income was
$60,879, and 15.4% of the population lived in poverty (U.S. Census
Bureau, 2010).
This study uses Census tracts to represent neighborhoods because
they are large enough to capture variation in low probability events
like child maltreatment referral, but small enough to approximate a
neighborhood and capture local effects that might be diluted or biased
downward if a larger, more heterogeneous spatial unit, such as zip
codes, were used (Ernst, 2001). Census tracts also take into account
population similarity to some extent as opposed to the convenience of
mail routes, and the U.S. Census Bureau makes a concerted effort
when constructing tract boundaries to consider the presence of physical
features, such as major highways that may dissect and define how res-
idents perceive their neighborhoods (U.S. Census Bureau, n.d.).
2.2. Measures
2.2.1. Dependent variables
The total number of child maltreatment referrals per Census tract in-
volving Black, Hispanic, and White children birth through 17 years were
downloaded from the Advancement Project's Healthy City website
(http://www.healthycity.org/) for all available years: 2006 through
2009. These data were provided to Healthy City by the Los Angeles
County Department of Children and Family Services (DCFS). We use
referrals rather than substantiated referrals to measure child maltreat-
ment because we believe they possess several advantages as an indica-
tor. First, research suggests that the distinction between substantiated
and unsubstantiated reports is not informative with respect to describ-
ing children's future outcomes, including the likelihood that they will be
re-referred to CPS. (Drake, Jonson-Reid, Way, & Chung, 2003; Kohl et al.,
2009). Additionally, Los Angeles County adopted a “differential re-
sponse” model of service delivery in 2004. Under this model, some mal-
treatment reports are not substantiated when families agree to accept
“voluntary” community-based services in lieu of formal CPS supervi-
sion. Thus, the substantiation decision is as much about case processing
as it is about verifying whether maltreatment actually occurred. Further
evidence that substantiation is a flawed indicator of the validity of child
maltreatment reports comes from Cross and Casanueva's (2009) analy-
sis of a national probability child welfare sample in which they found
that a significant portion (9%) of substantiation decisions were not
explained by evidence, harm or risk. While reports are also imperfect
indicators of child maltreatment because spurious allegations have not
been weeded out, for the reasons noted above we believe they are the
best available indicator.
Child's primary race/ethnicity is based on child welfare caseworker
records. DCFS staff code all referred children as either African
American (a.k.a. Black), American Indian/Alaskan Native, Asian/Pacific
Islander, Hispanic/Latino, White or Other. In so doing, DCFS conflates
race and Hispanic ethnicity, making it impossible to disentangle these
two constructs in the current study. It should also be noted that DCFS
practice does not require that workers base their records on families'
self-identified race or ethnicity, allowing for misidentification based
on workers' perceptions and assumptions. Because the total number
of referrals involving American Indian/Alaskan Native, Asian/Pacific
Islander, and Other children was too small to enable reliable estimates
at the Census tract level, we restricted our analysis to Black, Hispanic,
and White children only, examining the neighborhood characteristics
that predict maltreatment referral rates for each of these groups
separately.
The Los Angeles County DCFS used Environmental Systems Research
Institute's ArcGIS 9.2 program and the Thomas Brothers Transportation
Street Network Line database were used to geocode the home addresses
associated with all of the 2006–2009 child maltreatment reports that it
received for each racial/ethnic group of 0–17 year old children, and then
associated the address coordinates with Census tracts in Los Angeles
County to calculate the total number of child maltreatment reports
within each tract. For the current study, we calculated referral rates
per 1000 children in each racial/ethnic group by applying U.S. Census
2000 estimates of the population of Black, Hispanic and White children
0–17 years old residing in each tract using Census 2000 tract bound-
aries. It should be noted that the dependent variables in this study are
based on the number of child maltreatment reports, not the number
of children reported. Consequently, maltreatment risk may be slightly
overstated in Census tracts containing children who were referred to
DCFS multiple times during the study period, and it may be slightly
underestimated in Census tracts with maltreatment reports involving
multiple children (as might happen in the case of a report involving
several children from the same family). Because geocoding translated
the individual level data into neighborhood-level data, both Michigan
State University and New York University Offices for Protection of Re-
search Subjects determined that this research study did not involve
“human subjects” and therefore did not require Institutional Review
Board review or approval.
In this study, child maltreatment referral rates are measured as the
four-year average for 2006–2009 of the number of child maltreatment
referrals per Census tract for each race/ethnicity (Black, Hispanic, and
White) divided by the total number of child residents of that race/
ethnicity residing in the tract and then multiplied by 1000. For the
three racial/ethnic groups that we examined, tract-level child maltreat-
ment referral rates for Black children were much higher than for Hispanic
and White children. The average rate of Black child maltreatment
referrals for Los Angeles County Census tracts in 2006–2009 was 126
per 1000 Black children birth through 17 years (standard deviation:
225). The average rate of Hispanic child maltreatment referrals was 56
per 1000 Hispanic children birth through 17 years (standard deviation:
171). The average rate of White child maltreatment referrals was 64 per
1000 White children birth through 17 years (standard deviation: 104).
All three of these dependent variables have heavily skewed distributions
because CPS referral is a low base rate phenomenon and, thus, there are
 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
many Census tracts with zero referrals. To help mitigate against the non-
normality of the dependent variables biasing our final analyses, we trans-
formed the dependent variables. A series of potential transformations
(cubic, square, identity, square root, log 1/square root, inverse 1/square
and 1/cubic) were explored using Stata's “gladder” command, and a
square root transformation was identified as most optimal for all three
of the dependent variables and applied.
2.2.2. Independent and control variables
Our primary independent variable of interest, racial-ethnic heteroge-
neity (diversity), was measured using Blau's Diversity Index (Blau,
1977). It is the percentage of times, expressed as a ratio, in which two
people randomly selected from an area will differ by race/ethnicity.
The index is calculated by squaring the percent of each mutually exclu-
sive racial/ethnic group residing in an area expressed as a fraction, then
summing the squares and subtracting the total from 1.00. The following
racial/ethnic categories were included in the construction of this mea-
sure: a) Hispanic/Latinos of any race, b) non-Hispanic Whites of one
race, c) non-Hispanic African Americans of one race, d) non-Hispanic
American Indians and Alaska Natives of one race, e) non-Hispanic
Asians, Native Hawaiians, and other Pacific Islanders of one race,
f) non-Hispanics of some other race, and g) non-Hispanics of two or
more races. This index ranges from a theoretical value of 0, signaling
that all residents share the same race/ethnicity, to +1, which represents
extreme diversity.
We constructed this variable using, 2000 U.S. Census estimates.
While we explored using updated Census estimates from several
sources that more perfectly corresponded with the years that our
child maltreatment data was collected (e.g. American Community
Survey 2005–2009), we ultimately rejected these data as insufficiently
reliable for our analysis because they are based on samples rather
than a full census of the entire population. Such estimates tend to be
most reliable for large spatial units, while our analysis focuses on
racially/ethnically-specific child-only sub-populations residing within
small areal units of analysis (Census tracts).
Several demographic variables used as indicators of social disorgani-
zation in previous research and found to be related to child maltreat-
ment were included in the analysis as control variables. All of these
variables were also provided by, or constructed using, 2000 U.S. Census
estimates. To control for the degree of economic disadvantage, or con-
centrated poverty, within neighborhoods, we considered the percent of
residents in each tract with incomes below the Federal Poverty Line
(i.e. $17,050 for a family of four). Child care burden was measured by:
1) the ratio of children-to-adults (total number of children birth through
17 years old divided by the total number of adults 18 through 64 years
old) and 2) the elderly population (percentage of residents 65 years and
older). Residential instability was measured as the percentage of resi-
dents five years and older who did not live in their current home five
years previous, in other words those who had experienced a fairly
Table 1
Descriptive statistics for independent variables.
Variable M SD MIN MAX
Racial–ethnic heterogeneity 0.48 0.18 0 0.78
Immigrant concentration (%) 35.54 16.42 0 88.90
Concentrated poverty (%) 17.87 13.03 0 100
Child care burden:
Child/adult ratio 0.46 0.17 0 1.49
Elderly population (%) 10.07 5.90 0 100
Residential instability (%) 52.16 11.79 0 95.60
Housing stress:
Vacant housing (%) 4.36 4.09 0 100
Crowded homes (%) 17.13 15.57 0 100
Note. M = mean; SD = standard deviation; MAX = maximum; MIN = minimum;
U.S. = United States.
99
recent move. Immmigrant concentration reflects the percentage of resi-
dents who were born outside the United States. Lastly, housing stress
was measured in two ways: 1) vacant housing reflects the percentage
of housing units in a Census tract that were unoccupied, while
2) crowded homes is the percentage of housing units with more than
one and half occupants per room. Table 1 presents descriptive statistics
for the independent and control variables used in this study, and Table 2
presents a correlation matrix for all of the dependent, independent and
control variables in this study.
3. Data analysis procedures
This study uses spatial error regression models to examine how child
maltreatment referral rates are related to the racial and ethnic diversity
within neighborhoods as well as several other indicators or neighbor-
hood social disorganization. We selected this analytic approach because
exploratory data analysis revealed positive spatial autocorrelation in the
dependent variables, a frequent concern in neighborhood-level analyses.
3.1. Spatial autocorrelation
We conducted exploratory spatial autocorrelation analyses using the
Moran I statistic to assess the spatial dependence of the dependent var-
iables in each of our models described below. Moran I measures global
autocorrelation across all the spatial units in a data set. It has a range
of − 1 to 1, with a zero value indicating a completely random spatial
pattern, a negative value indicating perfect dispersion, and a positive
value indicating spatial correlation or clustering. All three of our depen-
dent variables have statistically significant, positive Moran I scores, sug-
gesting a spatial autocorrelation problem. For Black children, the Moran
I is 0.57, p b 0.001. For Hispanic children, the Moran I is 0.64, p b 0.001.
Finally, for White children, the Moran I is 0.40, p b 0.001.
Most existing neighborhood research on child maltreatment uses
standard ordinary least squares (OLS) regression techniques to analyze
the data despite the fact that contiguous neighborhoods tend to be more
similar to each other than distal neighborhoods, thereby violating the
OLS assumption of unit independence (Freisthler et al., 2006). This ten-
dency of variables of similar values to cluster together geographically is
known as positive spatial autocorrelation. In this study, it is indicated by
the presence of a pattern whereby neighborhoods with high rates of
child maltreatment referrals tend to be adjacent to each other and
neighborhoods with low rates of referral rates tend to be adjacent to
each other. Failure to employ spatial analytic techniques to assess and
control for this spatial effect would potentially result in correlated mea-
surement errors, biased parameter estimates, and increased possibility
of committing a Type I error (Haining, 2003).
3.2. Spatial error regression
To correct for spatial autocorrelation we used GeoDa 1.4.0 software
to specify generalized least squares (GLS) spatial error regression
models for Black, Hispanic, and White child maltreatment referral
rates. There are different ways to control for spatial autocorrelation de-
pending on the nature of the data. One common approach is to conduct
a spatial lag regression, however, this technique is more appropriate
when the spatial autocorrelation associated with the dependent
variable reflects a contagion effect. As we have no theoretical basis for
assuming that child maltreatment referrals are ‘catching’ from one
neighborhood to the next, we used a spatial error regression model in-
stead. This procedure is similar to a standard regression model, except
we include a spatially weighted error term, thus treating the spatial au-
tocorrelation as an error or ‘nuisance’. The spatial error regression
model used in this study can be expressed as:
yi ¼ xi β þ λwi ξi þ ei ;
100 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105

Table 2
Correlation matrix.

1 2 3 4 5 6 7 8 9 10 11

1. BlkRefRate 1
2. HispRefRate .03 1
3. WhtRefRate .15 .15 1
4. RaceEthHet −.08 −.02 −.06 1
5. ImmigConc .10 −.02 .05 −.09 1
6. %Poor .15 .05 .29 −.13 .55 1
7. Child:Adult .04 .03 .25 −.22 .28 .47 1
8. %Elderly −.02 −.05 −.15 .07 −.29 −.45 −.46 1
9. ResInstab −.05 −.05 −.02 −.10 −.19 −.35 .17 .27 1
10. VacHouse .09 .07 .12 −.08 −.05 .27 .04 −.09 −.23 1
11. CrwdHome .12 .03 .20 −.34 .76 .75 .59 −.51 −.22 .10 1

Note. All correlations at or above /0.05/ are significant at p ≤ .01.

1. BlkRefRate = Black 2006–09 child maltreatment referral rate; 2. HispRefRate = Hispanic/Latino 2006–09 child maltreatment referral rate; 3. WhtRefRate = White 2006–09 child
maltreatment referral rate; 4. RaceEthHet = racial–ethnic heterogeneity; 5. ImmigConc = immigrant concentration; 6. %Poor = poverty rate; 7. Child:Adult = child to adult ratio;
8. %Elderly = elderly population; 9. ResInstab = residential instability; 10. VacHouse = vacant housing; and 11. CrwdHome = crowded homes.

where y is the dependent variable, x is the independent variable and β is
the coefficient associated with the dependent variable. The error term
consists of two parts; e is a spatially uncorrelated error term, and ξ is
the spatially correlated error term for which λ specifies the strength
of the correlation between neighboring spatial units, and w is a
pre-specified spatial weights matrix that provides the structure of the
assumed spatial relationship (Ward & Gleditsch, 2008). We used a
rook's first order contiguity matrix to define which Census tracts quali-
fied as ‘neighbors’. This approach categorizes Census tracts that share a
boundary with each other (i.e. tracts that touch each other on at least
one edge) as ‘neighbors’.
3.3. Heteroskedasticity
Because population size is distributed unevenly across neighbor-
hoods creating heteroskedasticity, spatial models involving population
characteristics frequently have uneven error terms. Using maltreatment
referral rates rather than counts helps address this problem, but rates
based on spatial units with small populations remain vulnerable to
artificial elevation resulting from insufficient data as opposed to real
risk. To address the potential source of bias from this ‘small area
problem’ we adjusted each of our models with the square root of the
race/ethnicity-specific child population (birth to age 17) for that area
(Freisthler et al., 2006).
3.4. Analytic models
Our first analytic model regresses maltreatment referral rates for
Black children on neighborhood racial-ethnic heterogeneity and the
control variables. The second model regresses maltreatment referral
rates for Hispanic children on the same variables and the third model
regresses maltreatment referral rates for White children on these vari-
ables. GeoDa provides a pseudo-R-squared statistic that estimates the
amount of variance in the dependent variable that each model explains
(see Table 3), allowing us to compare the relative explanatory power of
our model for each of the three racial/ethnic groups we examine. The
Log Likelihood statistic also provides a measure of model fit, with larger
numbers indicating better model fit.
While transforming our dependent variables helped normalize their
distribution, there remained a large number of tracts with zero referrals
given that maltreatment allegations, particularly when parsed by race/
ethnicity, are a low base rate phenomenon. Therefore, we also ran

Table 3
Generalized least squares (GLS) spatial error regression models of 2006–2009 neighborhood maltreatment referral rates (N = 2052) in Los Angeles County for Black, Hispanic, and White
children.

Black Hispanic White

Referral rates Referral rates Referral rates

Raw Moran I 0.57⁎⁎⁎ 0.64⁎⁎⁎ 0.40⁎⁎⁎

Independent variables b se b se b se

Constant 13.9011⁎⁎⁎ 0.6450 10.3355⁎⁎⁎ 0.8094 7.5658 0.7827

Racial–ethnic heterogeneity 0.3964⁎ 0.1966 0.3520⁎⁎⁎ 0.0470 0.7610⁎⁎⁎ 0.0943
Immigrant concentration 0.0072 0.0042 0.0013 0.0009 −0.0086⁎⁎⁎ 0.0015
Concentrated poverty 0.0064 0.0036 0.0104⁎⁎⁎ 0.0025 0.0113⁎⁎⁎ 0.0029
Child care burden:
Child/adult ratio 0.3762 0.2680 0.4285⁎⁎⁎ 0.0732 0.5372⁎⁎⁎ 0.1184
Elderly population 0.0145 0.0077 0.0104⁎⁎⁎ 0.0025 0.0018 0.0029
Residential instability 0.0013 0.0025 0.0016⁎ 0.0007 −0.0041⁎⁎⁎ 0.0012
Housing stress:
Vacant housing 0.0543⁎⁎⁎ 0.0099 0.0095⁎⁎⁎ 0.0029 0.0165⁎⁎⁎ 0.0047
Crowded homes 0.0035 0.0053 0.0006 0.0012 0.0095⁎⁎ 0.0032
Lambda 0.4395⁎⁎⁎ 0.0281 0.4885⁎⁎⁎ 0.0268 0.4266⁎⁎⁎ 0.0284
Pseudo-R2 0.58 0.76 0.38
Log likelihood test −7891 −6915 −7082

Note.
⁎ p b .05.
⁎⁎ p b .01.
⁎⁎⁎ p b .001.
 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
several aspatial count models (Poisson, negative binomial, zero-inflated
Poisson and zero-inflated negative binomial) to confirm the results of
the spatial regressions.
4. Results
Descriptive statistics for the independent and control variables in
this study are presented in Table 1. During the four years spanning
2006 through 2009, the typical Los Angeles Census tract was racially/
ethnically heterogeneous. The average Diversity Index score indicates
that there was a 48% chance that two residents of the same neighbor-
hood (tract) chosen at random would be members of differing racial/
ethnic groups.
The poverty rate for the average Los Angeles County Census tract in
2000 was almost 18%, with much variability indicated by the high stan-
dard deviation and wide range of scores. Residential turnover was com-
monplace with slightly more than half the residents of the average tract
having moved during the past five years. Regarding child care burden,
there were roughly two adults for every child in the average Census
tract, and the percentage of residents 65 years or older was generally
low at just over ten percent for the average tract. There was also a
high concentration of immigrants. On average, slightly more than a
third of the residents in a typical Los Angeles County Census tract
were born outside the U.S., however their concentration varied substan-
tially from tracts composed entirely of native born residents to tracts
composed almost exclusively of immigrants (maximum: 89%).
Table 2 provides the simple correlations among all the variables used
in the study. As depicted, Black, Hispanic and White referral rates were
inversely related to racial–ethnic heterogeneity, although the size of
these correlations was small. Most of the remaining correlations are
significant but not overly large. The only notably high correlation is
the relationship between residential crowding and immigrant concen-
tration, which is 0.76.
4.1. Regression interpretation
Table 3 presents the results of the multi-variate spatial error regres-
sion models, their pseudo-R2 values, and the model fit statistics (log
likelihood test). The models for Black and Hispanic children explain
more than half of the observed neighborhood-level variability in child
maltreatment referral rates (58% and 76% respectively), while the
third model for White children explained slightly more than a third of
the variability in referral rates (38%).
Consistent with social disorganization theory and our hypotheses,
racial–ethnic heterogeneity predicted higher rates of maltreatment re-
ferrals for Black children. Vacant housing also emerged as a risk factor
for maltreatment referrals of Black children. However, contrary to social
disorganization theory, there was no statistically significant relationship
between maltreatment referral rates for Black children and poverty,
immigrant concentration, child care burden, residential instability, or
residential crowding.
For Hispanic children, the racial–ethnic heterogeneity as risk factor
hypothesis was also supported. Hispanic children residing in racially/
ethnically diverse neighborhoods were referred to child protection
authorities at higher rates than those living in more homogeneous
neighborhoods. Also consistent with social disorganization theory,
neighborhood impoverishment, residential churning, vacant housing,
and greater levels of child care burden as measured by the child-to-
adult ratio and the percentage of elderly residents predicted higher
maltreatment referral rates for Hispanic children. Contrary to social dis-
organization theory, however, neither residential crowding nor the con-
centration of immigrants in a neighborhood predicted Hispanic child
maltreatment referral rates.
As with Black and Hispanic children, the racial–ethnic heterogeneity
hypothesis was also supported for White children; on average, the more
racially/ethnically diverse the neighborhood, the higher the child
101
maltreatment referral rate for White children. Most of the other indica-
tors of social disorganization also predicted higher referral rates for
White children, specifically concentrated poverty, residential instability,
housing stress, and child care burden as measured by the child to adult
ratio (but not the elderly population). Interestingly, immigrant concen-
tration appears to function as a protective factor for White children,
with a higher concentration of immigrants predicting lower maltreat-
ment referral rates for White children.
4.2. Model confirmation
The statistically significant and positive relationship between our
key predictor of interest, racial–ethnic heterogeneity, and child maltreat-
ment referrals was confirmed by a series of aspatial count models
(poisson, negative binomial, zero-inflated Poisson, and zero-inflated
negative binomial). Of these, the zero-inflated negative binomial
model provided the best fit for our data, and the count portion of this
model resulted in roughly the same fit as the spatial models described
above. In the minority of instances when the coefficients differed with
respect to significance or direction, the spatial model usually reflected
a more conservative result.
5. Discussion
The presented research assessed the relationship between racial-
ethnic diversity (‘heterogeneity’) within neighborhoods and child
maltreatment referral rates for Black, Hispanic and White children. We
conducted spatial error regression analyses controlling for indicators
of social disorganization that have been previously associated with
community-level child maltreatment rates, namely impoverishment,
child care burden, residential instability, immigration concentration,
and housing stress. Our hypothesis was that racial–ethnic diversity
would predict higher neighborhood child maltreatment referral rates
for all three of the racial/ethnic groups studied.
Despite a well-established theoretical and empirical rationale for ex-
amining racial–ethnic heterogeneity as a predictor of neighborhood so-
cial disorganization (Osgood & Chambers, 2000; Sampson & Groves,
1989; Sunghoon & Choo, 2008), child maltreatment researchers have
typically avoided incorporating this particular construct into their re-
search, instead opting to explore the relationship between neighbor-
hood racial/ethnic composition and child maltreatment by examining
the concentration of specific racial/ethnic groups in neighborhoods
(e.g. the percentage of residents who are Black rather than the level of
overall diversity irrespective of the particular racial/ethnic groups).
Our results indicate that residence in a racial/ethnic diverse neigh-
borhood is a clear risk factor for child welfare system involvement.
Child maltreatment reporting rates were higher for Black, Hispanic
and White children under the age of 18 residing in racially/ethnically di-
verse neighborhoods. This is consistent with social disorganization the-
ory which posits that racially/ethnically homogeneous neighborhoods
tend to be more cohesive because neighbors share cultural norms and
are not divided by racism, and consequently these neighborhoods are
better able to enforce community values, including values around ap-
propriate parenting and child safety.
Low levels of social cohesion in diverse neighborhoods may have a
two-fold impact on maltreatment reporting: one representative of actu-
al maltreatment risk and the other based on misguided opinions about
unfamiliar parenting techniques. It is possible that parents in these
types of neighborhoods are more likely to experience social isolation
and are less able to count on neighbors to help with childcare or provide
other supports that can mitigate parenting stress, thereby leaving them
at greater risk of abusing or neglecting their children. Maltreatment re-
ports may also be higher in diverse neighborhoods due to a greater like-
lihood of false reporting. Residents may be more prone to misjudging or
coding parenting techniques used by other-raced neighbors as abusive
or neglectful. One would hope that living in a diverse neighborhood
102 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
would increase exposure to cultural differences, and thus result in un-
derstanding and tolerance. However, lacking cultural commonalities
may in fact contribute to suspicion of misunderstood behavior and
result in unfounded accusations of maltreatment.
Our finding that racial–ethnic diversity predicted child maltreat-
ment higher reporting rates for all three groups of children studied
can be interpreted as lending credence to social disorganization theory.
However, it is important to note that our study is exploratory and de-
scriptive in nature. It is not designed to test theory, and so we caution
the reader against making too much of the theoretical implications of
this finding, as there may be other mechanisms besides neighborhood
social organization that explain the observed link between diversity
and child maltreatment. For instance, there may be something endoge-
nous about people who live in diverse neighborhoods that makes them
more likely to engage in child abuse or neglect. If people tend to prefer
living in communities where their neighbors are ‘like’ them, then it
follows that many residents of diverse neighborhoods are not there by
choice but because they lack agency and resources to select a more
monoracial environment. The ‘tough times’ that restrict their ability to
move somewhere else may also contribute to higher levels of stress,
less access to resources to buffer this stress, and consequently a greater
propensity for maltreating their children.
We found that racial–ethnic diversity was the only indicator of
neighborhood disorganization that predicted higher maltreatment re-
ferral rates consistently for all three racial/ethnic groups of children
studied. Some social disorganization scholars have speculated that the
presence of large numbers of immigrants in a community may disrupt
cohesion in much the same way that racial–ethnic diversity is thought
to, but, in our study, immigrant concentration did not predict elevated
rates of reported child maltreatment for Black or Hispanic children;
and it actually functioned as a protective factor for White children. In
another California-based study, Freisthler, Gruenewald, Remer, Lery,
and Needell (2007) also found that the percent of immigrants failed to
predict higher rates of child maltreatment. It is possible that immigrant
concentration is predictive of actual child maltreatment as theorized,
but this relationship is masked in these studies because immigrants
are less inclined to report abusive or neglectful neighbors to CPS based
on a general distrust of authorities and lack of familiarity with the
American child protection system. Alternatively, immigrant concentra-
tion may be a poor indicator of social disorganization. There may in
fact be strong social supports within immigrant enclaves that help mit-
igate other maltreatment risk factors such that immigrant concentra-
tion is a proxy of social cohesion rather than disunity. This would help
explain why several studies of community violence, not just child
maltreatment research, also find no link between the proportion of im-
migrant or Latino residents and violence or victimization (Lee et al.,
2001; Martinez, 2002; Morenoff et al., 2001; Sampson & Raudenbush,
1999). As for our unique finding that immigrant concentration predicts
lower maltreatment reporting rates for White children, we can only
speculate that the large number of immigrants living in Los Angeles,
many of whom are undocumented, are particularly apprehensive
about passing judgment or engaging with authorities where it concerns
parenting techniques of the dominant culture. Future research should
establish the reliability of this finding through replication.
Social disorganization theory predicts that maltreatment referral
rates for all children will be higher in impoverished neighborhoods.
We found this to be true for Hispanic and White children, but not
Black children. Therefore, we join several other researchers in suggest-
ing that residence in a poor neighborhood matters differently for chil-
dren of different races/ethnicities. While Korbin et al. (1998) found
that tract-level poverty predicted maltreatment referral rates in pre-
dominantly White and predominantly Black tracts, they observed that
the relationship was much weaker in the Black tracts. Moreover, when
Drake and Pandy (1996) stratified Census tracts in Missouri both by ra-
cial composition and poverty rates, they found a pattern of ‘differential
sensitivity’ to poverty whereby White children living in poor
communities were more likely to be reported to CPS than Black children
living in poor communities, but White children living in non-poor com-
munities were less likely to be reported to CPS than Black children living
in non-poor communities. “Differential sensitivity” is also suggested by
Wulczyn, Gibbons, Snowden, and Lery (2013) study of the Black/White
foster placement gap in California counties, which found that racial
disproportionality in placement rates narrowed substantially when
analyses were restricted to counties with high rates of White poverty,
but this was not true for counties with high rates of Black poverty.
Our findings may reflect a ‘differential sensitivity’ to other indicators
of social disorganization, as well. While child care burden and residen-
tial instability predicted higher child maltreatment reporting rates for
White and Hispanic children, they did not for Black children. Korbin
et al. (1998) results were similar. They found that child care burden
and residential turnover predicted higher rates of confirmed child mal-
treatment reports in Census tracts with mostly White residents, but not
in tracts with mostly Black residents. Freisthler, Gruenewald, Remer,
Lery, and Needell (2007) likewise found that child care burden was
unrelated to child maltreatment rates for Black children. Korbin et al.
(1998) note that these findings are consistent with research documenting
high levels of “strength and solidarity” (social organization) in a number
of impoverished African American communities (p. 229).
It is also possible that Hispanic and White families are more depen-
dent on the availability of informal community resources and the stabil-
ity of their environment. Spearly and Lauderdale (1983) suggest that
this may be true for Latinos because cultural norms encourage them
to rely more upon neighbors for informal supports (e.g. child care) rath-
er than on formal support services that can be more easily replaced
upon moving to a new community. Another possibility is that the
most disadvantaged Blacks are those that are ‘stuck’ in disorganized
neighborhoods where residential stability is indicative of an inability
to ‘move up’ to less stressful and more supportive parenting environ-
ments rather than an indicator of social organization. Interestingly,
Freisthler, Gruenewald, Remer, Lery, and Needell (2007) findings
regarding residential instability and child maltreatment in central and
northern California were essentially the inverse of what we found in
Los Angeles; Census tracts where residents moved frequently had
higher rates of substantiated child maltreatment for Black children,
but not Hispanic and White children. Additional research is needed to
understand when and why the relationship between residential turn-
over and child welfare outcomes varies by child race/ethnicity.
5.1. Limitations
There are certain methodological limitations with this study that
warrant discussion. Whereas this study draws heavily on social disorga-
nization theory, the social processes highlighted by this theory, such as
collective efficacy and shared norms, are not directly measured; rather,
they are inferred from demographic patterns that have been linked to
these processes in prior research. As suggested above, there may be
some instances in which these measures are imperfect representations
of the underlying constructs that they are intended to represent.
Also, this study uses Census tracts as proxies for neighborhoods,
which may not coincide with residents' own definition of the communi-
ties in which they live, obscuring patterns of social interaction as they
actually occur. The tracts studied are limited to Los Angeles County,
California and thus findings may not be generalizable to other parts of
the country, particularly to more rural areas. Care should also be taken
in generalizing findings from this study to other urban areas given
that some aspects of Los Angeles neighborhood structure and popula-
tion dynamics are relatively unique. For instance, Los Angeles is the
most populous county in the nation by sizeable margins and unusually
diverse, with fewer non-Hispanic Whites (i.e. more ‘minorities’) than
any other county in the U.S. (Christie, 2007).
Moreover, while spatial error regression models offer the advantage
of controlling for spatial effects, one disadvantage of our analytic
 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
approach is that it does not produce standardized regression coeffi-
cients nor offer a reliable means of calculating the effect size of indepen-
dent variables. Thus, this study offers insight into whether the racial and
ethnic structure of neighborhoods is related to child maltreatment
reporting and the direction of these effects, but it does not tell us the
magnitude of the observed relationships. Additionally, this analysis is
based on child maltreatment reports rather than on children reported
(unduplicated reports), which may have introduced some ‘noise’, par-
ticularly in instances when multiple reports were made on the same in-
cident or when the same child was referred for multiple incidents.
Further, it should be noted that the current study does not include
Asian/Pacific Islanders and Native Americans because there were not
enough maltreatment reports involving these populations during the
study year to enable separate analyses. We agree with Freisthler,
Bruce, and Needell (2007) that attention should be paid to studying
these groups in future research.
Lastly, because this is an ecological study that does not include
resident-level variables, it is possible that some of the observed neigh-
borhood effects may be misattributed unmeasured individual-level ef-
fects. For instance, the observed relationship between neighborhood
diversity and higher child maltreatment reporting rates may reflect
the negative impact of living in a disorganized community as suggested,
but another possibility is that dysfunctional parents are simply more
likely to live in diverse communities. Until quite recently, it was not pos-
sible to control for spatial effects while using hierarchical regressions
techniques capable of modeling individual-level effects nested within
neighborhood effects; thus, even had we had access to population-
level data for the current study, we would have been forced to make a
trade-off between controlling spatial effects and hierarchical modeling
when conceptualizing the study. This is a significant limitation of the
current study as the majority of the variance in child maltreatment re-
ferral rates is at the individual rather than neighborhood level, and
neighborhood variance is best understood after taking that into account.
In the future, researchers with access to multi-level data may be able to
take advantage of advances in statistical software that allows them to
model how individual factors (e.g. race) interact with neighborhood
characteristics to influence child maltreatment while also controlling
for spatial effects.
Despite these limitations, this study makes several contributions to
the knowledge base on the relationship between neighborhood racial/
ethnic structure and the identification of child maltreatment. It also ex-
pands our understanding of the differential impact of social disorganiza-
tion on families of different races/ethnicities.
5.2. Policy and practice implications
Our study adds to the growing body of research that suggests neigh-
borhood environments influence child maltreatment risk, and thus
place-based interventions may be a promising avenue for reducing
child abuse and neglect. In particular, it adds to this literature by draw-
ing attention to the need for such interventions to consider neighbor-
hood diversity as a possible risk factor for child maltreatment. If
racial–ethnic heterogeneity increases child welfare system involvement
for the reasons that social disorganization theory suggests, it follows
that place-based child abuse prevention efforts should include activities
that focus on building trust, cohesion and shared norms about parenting
among neighbors from different cultural backgrounds. Our review of
the literature did not identify any evidence-based child maltreatment
programs with this objective, but we did find descriptions of some
promising approaches to building social cohesion in diverse communi-
ties. For instance, Hudson, Phillips, Ray, and Barnes (2007) describe
efforts of this kind underway in two racially/ethnically heterogeneous
neighborhoods in Britain. In Moss Side, Manchester these include a
Somali community organization establishing a series of summer
playgroups for children of various ages that were intentionally opened
to families throughout the community in order to facilitate interaction
103
between Somali and non-Somali parents. In North Tottenham, the Liv-
ing Under the Sun project uses organized cooking events as a means
of helping female Kosovan, Albanian, Turkish, Kurdish, Latin American,
Somali and West African neighbors and their children form positive re-
lationships across cultural lines. Our research suggests that activities
and programs like these that promote social inclusivity and integration
in diverse communities may also make neighborhood children safer
from abuse and neglect.
Our research also points indirectly to the importance of culturally re-
sponsive services. It is not sufficient that child abuse prevention services
exist within a neighborhood. To effectively combat child maltreatment,
it is critical that these services be viewed as accessible and welcoming
by all community members regardless of their race or ethnicity. Social
disorganization theory suggests that people in diverse communities
are more likely to isolate themselves from their neighbors thus contrib-
uting to a lack of cohesion and shared parental norms. Kissane (2010)
extends this idea, noting that people in diverse communities may also
isolate themselves from potentially helpful family and social support
services because they racially/ethnically code the providers as existing
to serve/help an ‘other’ racial/ethnic group than their own. Therefore,
to help ensure the efficacy of place-based child abuse prevention pro-
grams, implementers should take steps to ensure that community resi-
dents of all races/ethnicities feel comfortable accessing program
services. Toward this end, outreach efforts should be culturally respon-
sive. At a minimum, service staff should include people of the targeted
racial/ethnic group(s) and, when targeting Hispanics, at least some
staff should speak Spanish.
Our results also imply that place-based child maltreatment inter-
ventions should be selected and implemented with careful consider-
ation of the racial/ethnic composition of the targeted neighborhood
and who in fact is at risk in that particular community. While it is
well documented that children living in socially disorganized com-
munities are particularly at risk of becoming involved in the child
welfare system, our research finds significant variation in most of
the specific facets of disorganization associated with maltreatment
of children of different races/ethnicities. Thus we join Freisthler,
Gruenewald, Remer, Lery, and Needell (2007) in strongly cautioning
against a one size-fits-all approach to designing place-based child
abuse prevention programs.
Current policy efforts are highlighting government-sponsored pro-
grams such as home visiting and promising neighborhood programs,
many of which have child abuse prevention goals and use a place-
based approach to delivering services. These efforts require a compre-
hensive needs assessment in order to identify at-risk communities and
design services based on those characteristics of the community. In
light of our finding that the community characteristics associated with
child maltreatment vary substantially for children of different races/
ethnicities, we suggest that as these programs are being developed
and duplicated, the government should mandate that the community
needs assessment parse data according to race and ethnicity. In order
to strengthen communities, we need to dig a little deeper and identify
exactly who in the community is at risk, so that services and programs
are selected according to the particular risk and protective factors asso-
ciated with these populations.
Federal and local governments increasingly require grantees to im-
plement evidence-based social service programs. This involves grantees
establishing that credible research evidence exists to support a particu-
lar intervention program. We believe that the time has come to require
that publically funded services establish not merely that their program
works (for some people, somewhere), but that there is evidence that
it works for the particular racial/ethnic group(s) with the greatest
need. As our research suggests, different races and ethnicities experi-
ence different risk and protective factors. Just because an intervention
has been rigorously tested and found to reduce child welfare involve-
ment or prevent maltreatment for one group does not mean that it
will do so for another. Without this attention to human diversity, the
104 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105
evidence-based practice movement that is currently so en vogue runs
the risk of inadvertently becoming a tool of cultural hegemony.
In conclusion, the current study strongly suggests that the racial/
ethnic structure of neighborhoods should be considered in community
assessments of child maltreatment and in implementing place-based
child maltreatment prevention approaches. Maltreatment risk differs
for racial/ethnic groups according to the overall diversity of the commu-
nities in which they live. Further, our efforts to understand the likeli-
hood of child welfare referral among different racial/ethnic groups
requires distinguishing the respective influences of poverty, resident
turnover, child care burden, housing stress and the proportion of immi-
grants in neighborhoods. Our results corroborate a small but growing
body of research indicating that characteristics of place matter very
differently depending on one's race/ethnicity.
References
Blau, P.M. (1977). Inequality and heterogeneity. New York, NY: Free Press.
Christie, L. (August 9, 2007). The nation's most ethnically diverse counties. CNNMoney.
(Retrieved on March 8, 2014 from http://money.cnn.com/2007/08/08/real_estate/
most_diverse_counties/).
Coulton, C. J., Crampton, D. S., Irwin, M., Spilsbury, J. C., & Korbin, J. E. (2007). How neigh-
borhoods influence child maltreatment: A review of the literature and alternative
pathways. Child Abuse and Neglect, 31, 1117–1142.
Coulton, C. J., & Korbin, J. (1995). Community level factors and child maltreatment rates.
Child Development, 66(5), 1262–1276.
Coulton, C. J., Korbin, J. E., & Su, M. (1999). Neighborhoods and child maltreatment: A
multi-level study. Child Abuse & Neglect, 23(11), 1019–1040.
Cross, T. P., & Casanueva, C. (2009). Caseworker judgments and substantiation. Child
Maltreatment, 14(1), 38–52.
Deccio, G., Horner, W. C., & Wilson, D. (1994). High-risk neighborhoods and high-risk
families: Replication research related to the human ecology of child maltreatment.
Journal of Social Service Research, 18(3/4), 123–137.
Drake, B., Jolley, J., Lanier, P., Fluke, J., & Barth, R. P. (2011). Racial bias in child protection?
A comparison of competing explanations using national data. Pediatrics, 127(3),
471–478.
Drake, B., Jonson-Reid, M., Way, I., & Chung, S. (2003). Substantiation and recidivism. Child
Maltreatment, 8(4), 248–260.
Drake, B., Lee, S. M., & Jonson-Reid, M. (2009). Race and child maltreatment reporting: Are
blacks overrepresented? Children and Youth Services Review, 31, 309–316.
Drake, B., & Pandy, S. (1996). Understanding the relationship between neighborhood
poverty and specific types of child maltreatment. Child Abuse and Neglect, 20(11),
1003–1018.
Drake, B., & Rank, M. R. (2009). The racial divide among American children in poverty:
Reassessing the importance of neighborhood. Children and Youth Services Review,
31, 1264–1271.
Ernst, J. S. (2001). Community-level factors and child maltreatment in a suburban county.
Social Work Research, 25, 133–142.
Fluke, J.D., Jones-Harden, B., Jenkins, M., & Ruehrdanz, A. (2010). Research synthesis on
child welfare disproportionality and disparities. Washington, DC: American Humane
Association and Center for the Study of Social Policy.
Fluke, J.D., Yuan, Y. T., Hedderson, J., & Curtis, P. A. (2003). Disproportionate representa-
tion of race and ethnicity in child maltreatment: Investigation and victimization.
Child and Youth Services Review, 25, 359–373.
Freisthler, B. (2004). A spatial analysis of social disorganization, alcohol access, and rates
of child maltreatment in neighborhoods. Children and Youth Services Review, 26,
307–323.
Freisthler, B., Bruce, E., & Needell, B. (2007). Understanding the geospatial relationship of
neighborhood characteristics and rates of maltreatment for black, Hispanic and white
children. Social Work, 52(1), 7–16.
Freisthler, B., Gruenewald, P. J., Remer, L. G., Lery, B., & Needell, B. (2007). Exploring the
spatial dynamics of alcohol outlets and Child Protective Services referrals, substanti-
ations, and foster care entries. Child Maltreatment, 12, 114–124.
Freisthler, B., Gruenewald, P. J., Ring, L., & LaScala, E. A. (2008). An ecological assessment
of the population and environmental correlates of childhood accident, assault and
child abuse injuries. Alcoholism, Clinical and Experimental Research, 32, 1969–1975.
Freisthler, B., Merritt, D. H., & LaScala, E. A. (2006). Understanding the ecology of child
maltreatment: A review of the literature and directions for future research. Child
Maltreatment, 11(3), 263–280.
Freisthler, B., Midanik, L. T., & Gruenewald, P. J. (2004). Alcohol outlets and child physical
abuse and neglect: Applying routine activities theory to the study of child maltreat-
ment. Journal of Studies on Alcohol, 65, 586–592.
Freisthler, B., Needell, B., & Gruenewald, P. (2005). Is the physical availability of alcohol
and illicit drugs related to neighborhood rates of child maltreatment? Child Abuse &
Neglect, 29(9), 1049–1060.
Gillham, B., Tanner, G., Cheyne, B., Freeman, I., Rooney, M., & Lambie, A. (1998). Unem-
ployment rates, single parent density, and indices of child poverty: Their relationship
to different categories of child abuse and neglect. Child Abuse and Neglect, 22, 79–90.
Greenman, E., Bodovski, K., & Reed, K. (2011). Neighborhood characteristics, parental
practices and children's math achievement in elementary school. Social Science
Research, 40, 1434–1444.
Guterman, N.B., Lee, S. J., Taylor, C. A., & Rathouz, P. J. (2010). Parental perceptions of
neighborhood processes, stress, personal control and risk for physical child abuse
and neglect. Child Abuse & Neglect, 33, 897–906.
Haining, R. P. (2003). Spatial Data Analysis: Theory and Practice. Cambridge: Cambridge
University Press.
Hill, R. B. (2006). Synthesis of research on disproportionality in child welfare: An update. The
Casey/Center for the Study of Social Policy Alliance for Racial Equity. (Retrieved from
http://www.racemattersconsortium.org/docs/BobHillPaper_FINAL.pdf).
Hudson, M., Phillips, J., Ray, K., & Barnes, H. (2007). Social cohesion in diverse communities. Jo-
seph Rowntree Foundation. (Retrieved on March 8, 2014 from http://www.google.com/
url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CCoQFjAA&url=http%3A%
2F%2Fwww.jrf.org.uk%2Fsites%2Ffiles%2Fjrf%2F2036-social-cohesion-communities.
pdf&ei=fpQbU97ZKePz2QX7vICYCA&usg=AFQjCNEnWKAWKMR5C9wXRFu_
6b12NzFA9g&sig2=4tq37GS310PlWi0aKrNGaA).
Kasarda, J.D., & Janowitz, M. (1974). Community attachment in mass society. American So-
ciological Review, 39, 328–339.
Kim, H., Chenot, D., & Ji, J. (2011). Racial/ethnic disparity in child welfare systems: A lon-
gitudinal study utilizing the Disparity Index (DI). Children and Youth Services Review,
33, 1234–1244.
Kissane, R. J. (2010). “We call it the badlands”: How social-spatial geographies influence
social service use. Social Service Review, 84(1), 3–28.
Klein, S. (2011). The availability of neighborhood early care and education resources and
the maltreatment of young children. Child Maltreatment, 16(4), 299–310.
Kohl, P. L., Jonson-Reid, M., & Drake, B. (2009). Time to leave substantiation behind:
Findings from a national probability study. Child Maltreatment, 14(1), 17–26.
Korbin, J. E., Coulton, C. J., Chard, S., Platt-Houston, C., & Su, M. (1998). Impoverishment
and child maltreatment in African American and European American neighborhoods.
Development and Psychopathology, 10, 215–233.
Krishnan, V., & Morrison, K. B. (1995). An ecological model of child maltreatment in a
Canadian province. Child Abuse and Neglect, 19(1), 101–113.
Lee, M. T., Martinez, R., & Rosenfeld, R. (2001). Does immigration increase homicide?
Negative evidence from three border cities. Sociological Quarterly, 42, 559–580.
Martinez, R. (2002). Latino homicide: Immigration, violence and community. New York:
Routledge.
Massey, D. S. (2004). Segregation and stratification: A biosocial perspective. Du Bois
Review, 1(1), 7–25.
Merritt, D. H. (2009). Child abuse potential: Correlates with child maltreatment rates and
structural measures of neighborhoods. Children and Youth Services Review, 31(8),
927–934.
Molner, B. E., Buka, S. L., Brennan, R. T., Holton, J. K., & Earls, F. (2003). A multilevel
study of neighborhoods and parent-to-child physical aggression: Results from
the Project on Human Development in Chicago Neighborhoods. Child
Maltreatment, 8, 84–97.
Morenoff, J.D., Sampson, R. J., & Raudenbush, S. W. (2001). Neighborhood inequality, col-
lective efficacy, and the spatial dynamics of urban violence. Criminology, 39, 517–559.
Needell, B., Brookhart, M.A., & Lee, S. (2003). Black children and foster care placement in
California. Children and Youth Services Review, 25, 393–408.
Nielsen, A. L., Lee, M. T., & Martinez, R. (2005). Integrating race, place and motive
in social disorganization theory: Lessons from a comparison of Black and Lati-
no homicide types in two immigrant destination cities. Criminology, 43(3),
837–872.
Osgood, D. W., & Chambers, J. M. (2000). Social disorganization outside the metropolis:
An analysis of rural youth violence. Criminology, 38, 81–116.
Putnam, Robert D. (2007). E Pluribus Unum: Diversity and community in the twenty-first
century the 2006 Johan Skytte Prize Lecture. Scandinavian Political Studies, 30(2),
137–174.
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals. American
Sociological Review, 15, 351–357.
Sampson, R. J. (1988). Local friendship ties and community attachment in mass society: A
multilevel systemic model. American sociological review, 53, 766–779.
Sampson, R. J. (2001). How do communities undergird or undermine human develop-
ment? Relevant contexts and social mechanisms. In A. Booth, & A.C. Crouter (Eds.),
Does it take a village? Community effects on children, adolescents and families
(pp. 3–30). Mahwah, NJ: Lawrence Erlbaum.
Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing social dis-
organization theory. American Journal of Sociology, 94, 775–802.
Sampson, R. J., & Raudenbush, S. W. (1999). Systematic social observation of public
spaces: A new look at disorder in urban neighborhoods. American Journal of Sociology,
105, 603–651.
Shaw, C. R., & McKay, H. (1969 [1942]]). Juvenile delinquency and urban areas. Chicago:
University of Chicago Press.
Spearly, J. L., & Lauderdale, M. (1983). Community characteristics and ethnicity in the pre-
diction of child maltreatment rates. Child Abuse & Neglect, 7(1), 91–105.
Stoltzfus, E. (2005). Race/ethnic and child welfare. Washington, DC: Congressional
Research Services.
Sunghoon, R., & Choo, T. M. (2008). Looking inside Zone V: Testing social disorganization
theory in suburban areas. Western Criminology Review, 9(1), 1–16.
Thomas, P.M. (2011). Theoretical articulation on immigration and crime. Homicide
Studies, 15(4), 382–403.
U.S. Census Bureau (2010). Table 66. Families by type, race, and Hispanic origin: 2010.
(Retrieved on April 19, 2013 from http://www.census.gov/compendia/statab/2012/
tables/12 s0066.pdf).
United States Department of Health and Human Services, Administration for Children and
Families, Administration on Children, & Youth and Families, Children's Bureau
[USDHHS] (2011). Child maltreatment. (Retrieved on April 21, 2013 from b https://
www.acf.hhs.gov/sites/default/files/cb/cm11.pdf).
 S. Klein, D.H. Merritt / Children and Youth Services Review 41 (2014) 95–105 105

Ward, M.D., & Gleditsch, K. S. (2008). Spatial regression models. Thousand Oaks, CA:
Sage.
Wilson, W. J. (1987). The truly disadvantaged: The inner city, the underclass, and public
policy. Chicago: University of Chicago.
Wilson, W. J. (1996). When work disappears: The world of the new urban poor. New York:
Vintage Books.
Wulczyn, F., Gibbons, R., Snowden, L., & Lery, B. (2013). Poverty, social disadvantage, and
the black/white placement gap. Children and Youth Services Review, 35, 65–74.
Wulczyn, F., & Lery, B. (2007). Racial disparity in foster care admissions. Chapin Hall:
University of Chicago.
Zuravin, S. J. (1986). Residential density and urban child maltreatment: An aggregate
analysis. Journal of Family Violence, 1(4), 307–322.