G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS Child Abuse & Neglect xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Child Abuse & Neglect
Research article
Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density Kathryn Maguire-Jack a,∗ , Paul Lanier b , Michelle Johnson-Motoyama c , Hannah Welch b , Michael Dineen d a b c d
The Ohio State University College of Social Work, 325B Stillman Hall, 1947 N College Road, Columbus, OH 43210, USA University of North Carolina Chapel Hill, 325 Pittsboro Street, CB #3550, Chapel Hill, NC 27599, USA University of Kansas School of Social Welfare, 311 Twente Hall, 1545 Lilac Lane, Lawrence, KS 66045, USA National Data Archive for Child Abuse and Neglect (NDACAN), Cornell University, Ithaca, NY 14853, USA
a r t i c l e
i n f o
Article history: Received 23 March 2015 Received in revised form 25 May 2015 Accepted 27 May 2015 Available online xxx Keywords: Child maltreatment Spatial Hierarchical linear modeling Geographic variation Poverty Population density
a b s t r a c t There are documented disparities in the rates at which black children come into contact with the child welfare system in the United States compared to white children. A great deal of research has proliferated aimed at understanding whether systematic biases or differential rates of risk among different groups drive these disparities (Drake et al., 2011). In the current study, county rates of maltreatment disparity are compared across the United States and examined in relation to rates of poverty disparity as well as population density. Specifically, using hierarchical linear modeling with a spatially lagged dependent variable, the current study examined data from the National Child Abuse and Neglect Data System (NCANDS) and found that poverty disparities were associated with rates of maltreatment disparities, and densely populated metropolitan counties tended to have the greatest levels of maltreatment disparity for both black and Hispanic children. A significant curvilinear relationship was also observed between these variables, such that in addition to the most densely populated counties, the most sparsely populated counties also tended to have higher rates of maltreatment disparity for black and Hispanic children. © 2015 Elsevier Ltd. All rights reserved.
Introduction The United States child welfare system (CWS) has a history of differential treatment for children of different races. In the past, minority children were excluded from the CWS entirely, with black children initially excluded from the developing foster care system (Roberts, 2002). However, black and Hispanic children are currently represented in the CWS at rates that are different than their representation in the general population. In 2013, black children made up 15% and Hispanic children made up 23% of the total child population, but they were represented among child maltreatment victims at proportions of 21% and 22%, respectively (United States Census Bureau, 2014a; United States Department of Health and Human Services, 2015). Additionally, a recent study of the cumulative prevalence of confirmed child maltreatment by age 18 years estimates that black children have a cumulative prevalence of 20.9%, while the rates for Hispanic and white children are 13.0% and 10.7%, respectively (Wildeman et al., 2014).
∗ Corresponding author. http://dx.doi.org/10.1016/j.chiabu.2015.05.020 0145-2134/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
2
An abundance of research has documented associations between child maltreatment and negative long-term outcomes that last into adulthood. Children who are maltreated are more likely to struggle with behavioral problems (Felitti et al., 1998; Repetti, Taylor, & Seeman, 2002), mental health conditions (Holmes & Sammel, 2002), increased risk for criminal behavior and violent behavior (Fang & Corso, 2007; Widom & Maxfield, 2001), lower health-related quality of life (Corso, Edwards, Fang, & Mercy, 2008; Lanier, Kohl, Raghavan, & Auslander, 2015) and lower long-term economic well-being (Currie & Widom, 2010). Given the disproportional representation of black and Hispanic children among child maltreatment victims and the significant sequelae for victims, it is essential that we understand why racial disparities occur. If overrepresented children are not involved in this system as a result of need, but rather, systematic bias on the behalf of CPS workers, changes must be made immediately to reverse this trend. If, however, this phenomenon results from disproportionate risk of child maltreatment due to other inequalities within our society, then the problem is not that ethnic minority children are overserved by CPS, but rather, that they are maltreated at a higher rate than their non-minority peers. Under the latter scenario, intervention needs to occur prior to CPS involvement. In addition to variation by child/race ethnicity, child maltreatment rates vary geographically (Ernst, 2001; Fryer & Miyoshi, 1995; Paulsen, 2003; Vinson & Baldry, 1999; Zielinski & Bradshaw, 2006). At the county level, researchers have found associations between a number of poverty-related (Albert & Barth, 1996; Freisthler & Weiss, 2008; Hon-Yei, 1989; Spearly & Lauderdale, 1983) and substance use-related (Albert & Barth, 1996; Freisthler & Weiss, 2008) factors and child maltreatment rates, with important differences when comparing urban, suburban, and rural areas (Albert & Barth, 1996). These areas differ in terms of service accessibility and availability, social connections, and poverty rates (Slovak & Carlson, 2009), which may have important implications for rates of racial disparity in child maltreatment. To the extent that ethnic minority families are concentrated in counties with higher levels of poverty and lower access to resources that address these challenges, they may be at an increased risk for child maltreatment; thus contributing to the racial disparity in child maltreatment. For example, in a study of aboriginal children in Canada, it was found that race and ethnicity of individual families was not driving differential placement rates, but rather, certain characteristics of agencies in areas with high concentrations of aboriginal children, which the authors attributed to a lack of resources for aboriginal children (Fluke, Chabot, Fallon, MacLaurin, & Blackstock, 2010). In the U.S., Hispanic families are significantly more likely to live in highly segregated communities that have higher rates of social isolation and greater resource deprivation (Hipp & Yates, 2011; Lee, 2000; Parker & McCall, 1999). Further, neighborhoods that are highly segregated with a high concentration of black residents are less likely to have access to basic needs assistance, mental health and substance abuse providers, and employment services (Allard, 2009). Therefore, the disproportional representation of ethnic minority children in the U.S. CWS may be partly explained by disproportional residence in high poverty geographic areas. Theories of Disparity We use the term “disparity ratio (DR)” to refer to the ratio of the rate of a phenomenon in one racial group compared to the rate in another racial group and “disparity” to refer to the extent to which the ratio is not equal to 1 (Myers, 2010). It should be noted that while some researchers have used the term “disparity” to refer explicitly to unequal treatment of a minority group (Hill, 2006), we are not referring to that definition of the term. We use “disparity” to refer simply to the existence of differences in rates of phenomena, and do not assume that the cause of such difference is due to unequal treatment. Black–White Disparity Among researchers and advocates, there have been three major arguments to explain the overrepresentation of black children in the CWS. All three of these theories are based on the hypothesis that the long and complicated history of racial discrimination in the United States has had profound and lasting impacts on families, which contribute to disparity. The key difference between these theories lies in the level at which discrimination occurs and the mechanism through which it contributes to disparity. The first hypothesis is that racial bias at the individual level causes caseworkers to interpret observations differently for black children compared to white children (Chand, 2000; Webb, Maddocks, & Bongilli, 2002). In this scenario, caseworkers are thought to be more likely to substantiate maltreatment allegations and remove a child into out-of-home care when the child is black compared to a white child. A second theory, proposes that institutional racism has created policies and procedures for mandated reporters, social service caseworkers, and family court judges that are inherently biased against black children and families (Ards, Myers, Malkis, Sugrue, & Zhou, 2003; Hill, 2004). The third and final explanation is that current and historic discrimination has put black families at a significant disadvantage economically, which increases their risks for maltreatment (e.g. poverty, single parenthood, substance use), and as a result, black families have higher rates of maltreatment (Bartholet, Wulczyn, Barth, & Lederman, 2011; Drake, Jolley, & Lanier, 2011; Font, Berger, & Slack, 2012; Lanier, Maguire-Jack, Walsh, Drake, & Hubel, 2014). Though official statistics show that black residents of the United States are much more likely to live below the poverty level (approximately 27% of blacks compared to 10% of whites; U.S. Census Bureau, 2014b), such numbers significantly understate the level of socioeconomic disadvantage black families experience. Black children are roughly three times more likely to be poor, but are up to 14 times more likely to live in high childhood poverty neighborhoods compared to white children (Drake & Rand, 2009). Further, blacks have roughly 1/10 the assets of whites in the United States (Kochhat, 2015). Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
3
Despite higher rates of referrals overall regarding black children, prior research has shown that white children in higher poverty areas are more likely than black children to be the subject of a child maltreatment report (Drake, Lee, & JonsonReid, 2009). Drake and Rand (2009) proposed a complementary theory to the theories of racial disparity, referred to as the theory of differential assortment. This theory posits that given higher rates of wealth and assets and relatively unlimited mobility among white families who have not experienced the same historic forms of discrimination faced by black families (e.g. redlining), whites are much less likely to live in areas of high rates of poverty. When white families do live in highly impoverished areas, they are likely to be there due to a multitude of risk factors (e.g., substance abuse, mental health issues) when compared to black families, who are more likely to live in such areas as a result of racial segregation, regardless of individual risks. Therefore, it is expected that white families in poverty-stricken communities should have higher rates of child maltreatment reports compared to black families in similar communities, by virtue of differential levels of risk.
Hispanic–White Disparity Despite local variation in state and county estimates of disparity among Hispanics yielding observations of both overrepresentation and underrepresentation (Ards, Myers, Malkis, Sugrue, & Zhou, 2003; Church, Gross, & Baldwin, 2005; Dettlaff et al., 2011; Needell et al., 2012), theories regarding the disproportional representation of Hispanic children in the CWS have predominately focused on explaining the underrepresentation that has been observed in child maltreatment at the national level (Sedlak, 2010; United States Department of Health and Human Services, 2015). These theories have focused on trends that have been described as the “Hispanic health paradox,” as well as the differential surveillance of Hispanic families, particularly among families with immigrant members. Health paradoxes refer to patterns of morbidity and/or mortality for a particular group that is at odds with what would be expected given the group’s socioeconomic profile (Acevedo-Garcia & Bates, 2008). Such trends have been observed for certain infant and adult health outcomes, whereby Hispanics, particularly immigrants, often demonstrate better health when compared to their white and non-immigrant counterparts of similar socioeconomic position (Franzini, Ribble, & Keddie, 2001; Hayes-Bautista, 2002; Jasso, Massey, Rosenzweig, & Smith, 2004; Palloni & Arias, 2004; Palloni & Morenoff, 2001). The presence of health advantages among Hispanics and immigrants, despite socioeconomic disadvantage, has been attributed widely to protective social and cultural dynamics (e.g., familism, religiosity, salutary health behavior). This social-cultural explanation is often presented in association with an acculturation hypothesis that predicts the deterioration of health among Hispanics over time and across generations in the U.S. (i.e., health assimilation) (see Acevedo-Garcia & Bates, 2008). Another explanation for the health advantages among immigrants is referred to as the “healthy immigrant effect,” which posits that those who are able to migrate have certain health advantages that provide them with physical and financial abilities to migrate (McDonald & Kennedy, 2004). Recent studies arising from population-level linkages of child welfare and vital statistics data in California have revealed disparities in CWS involvement that are consistent with the Hispanic health paradox, whereby Hispanic children of foreignborn and U.S.-born Hispanics are less likely to be involved with the CWS when compared to their white counterparts after adjusting for socioeconomic position (Putnam-Hornstein, Needell, King, & Johnson-Motoyama, 2013). Studies have also identified the significant challenges that Hispanic immigrant caregivers in lower-income households face in meeting their children’s basic needs for food, medical care, and safe supervision (Johnson-Motoyama et al., 2014). Despite these risks, infants of Hispanic foreign-born mothers are consistently less likely to become involved with the CWS when compared to infants of Hispanic U.S.-born mothers across ethnic subgroups (e.g., Mexican, Central American) (Johnson-Motoyama et al., 2014). A second theory regarding the underrepresentation of Hispanics in national estimates of child maltreatment victimization posits underreporting as a result of differential surveillance (Putnam-Hornstein et al., 2013). The differential surveillance of maltreatment among Hispanic families has been attributed to differential exposure to mandated reporters in social, health, and child care settings due to lower utilization of these services (Flores et al., 2002; Karoly & Gonzalez, 2011), particularly among undocumented families whose access to health and social assistance are typically restricted by policies and practices at the federal level and in many states and localities. Among legal permanent residents, access to health and social programs typically require waiting periods, which may influence exposure to mandated reporters. Moreover, national and local evaluations suggest that the eligibility restrictions embedded in anti-immigrant policy initiatives successfully reduce the participation of immigrant families in public programs, regardless of legal status (Borjas, 2011; Fix & Passel, 2002; Zimmerman & Fix, 1998), as policies that negatively impact immigrant caregivers typically have downward effects on citizen children’s participation in public programs (Fix & Zimmerman, 2001; Seiber, 2013). Such trends may reduce the number and types of contacts that Hispanic children have with mandated reporters. Underrepresentation may also arise from the differential mobility of immigrants. For example, many Mexican and Central American immigrants cross the U.S.–Mexico border throughout the year in seasonal patterns that correspond with U.S. labor demand (Dockterman, 2011; Passel & Cohn, 2009). Hispanic children, particularly those with foreign-born mothers, may be at lower risk for CWS involvement when compared to whites and children of native-born Hispanic mothers due to migration patterns that limit exposure to mandated reporters. Finally, some researchers suggest that paradoxical patterns might be due to inconsistent ways of defining Hispanic identity (Franzini, Riblle, & Keddie, 2001; Jasso et al., 2004; Palloni & Morenoff, 2001). Given the limited ability of most studies to Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
4
account for all of these possible explanations, the question of whether a Hispanic health paradox exists with regard to child maltreatment is far from settled. Background Drake et al. (2011) sought to test some of these explanations for disparity by comparing national rates of child maltreatment reports from the National Child Abuse and Neglect Data System (NCANDS). They compared disparity in child maltreatment to disparity in poverty as a risk factor for maltreatment, as well as disparity in a number of other negative outcomes for children that were deemed to be less susceptible to bias including infant mortality, low birth weight, premature birth, infant accidents, and infant homicides. The authors found that the black–white disparity in child maltreatment was very similar to the rates of disparity in poverty as well as all of the other negative outcomes for children. As a result, they concluded that child maltreatment disparity was likely the result of increased risk for maltreatment among black families, and not a function of reporter bias. Although the Hispanic–white disparity ratio for poverty was similar to that for black children compared with white children, data suggested that the presence of protective moderating factors offset the relationship between poverty and poor outcomes for Hispanic children, lending support to the notion of a Hispanic health paradox in child maltreatment (Drake et al., 2011). In an extension of this work, Lanier et al. (2014) conducted a state by state comparison of child maltreatment risk measures (e.g., rates of single mothers, rates of births to teen mothers) and found child maltreatment disparity to be highest in states with high rates of poverty disparity, providing additional support for the findings by Drake et al. (2011). On average, the level of maltreatment disparity for Hispanics was lower than that of blacks; however, the relationships of maltreatment to poverty remained consistent across the two groups to suggest the presence of protective factors that potentially mitigate the risk of maltreatment among Hispanic children (Lanier et al., 2014). Given the variation in maltreatment disparity that has been observed at the local level, particularly among Hispanic children, the current study delves deeper into these relationships by examining the trends of child maltreatment and poverty disparity at the county level. Specifically, we examine three questions: (1) Is there a relationship between poverty and maltreatment such that counties with higher disparity in poverty also have higher maltreatment disparity?; (2) Is there a relationship between maltreatment disparity and county population geography (i.e., rural vs. urban)?; and (3) Do these relationships differ for black–white disparity when compared to Hispanic–white disparity? A related study examined the relationship between county-level poverty and county-level racial disparities in foster care placement between black and white children (Wulczyn, Gibbons, Snowden, & Lery, 2013). In that study, the authors found that areas with higher levels of white poverty had a smaller gap between black and white rates of foster care placement; and that black poverty rates largely did not explain the difference in the rates. Our study builds on this prior work in three important ways (1) by focusing on an earlier step in CPS decision-making, that is, the rate of child maltreatment reports rather than foster care placements; (2) by examining disparities in poverty rates by race, rather than examining the rate for each race separately; and (3) by examining these relationships in counties in 48 states as well as Washington DC rather than 16 states. In doing so, we seek to advance knowledge about how to effectively reduce racial disparities in maltreatment. Method This study examined the association between three county-level indicators for white, black, and Hispanic children: maltreatment, poverty, and geography. We used two approaches to explore this relationship. First, we used mapping techniques to visually observe county-level variation in the indicators across the United States. Second, we used multivariate regression analyses to statistically test for observed relationships. Non-Hispanic white, non-Hispanic black, and Hispanic groups (without reference to race) were included because data for these groups were available in most states for the child indicators used in this study and were of key interest to expand on prior research, theory, and practice. Area Definitions County was chosen as the geographic unit of interest in this study. Prior work from Lanier et al. (2014) suggested statelevel variation in maltreatment DR, which prompted our interest in examining whether the large geographic unit of a state was masking important variation within. We selected county as the unit of analysis because it was the smallest level of geography available within the child maltreatment dataset. Data Sources and Variables Three data sources were linked at the county level using the Federal Information Processing Standard (FIPS), a unique spatial code. First, the unduplicated count of children who were the subject of a child maltreatment report was obtained from the 2011 NCANDS Child File. NCANDS data contain all reports of maltreatment to CPS for all states (and Washington, DC) except Oregon and Pennsylvania, which were excluded from the analysis. Oregon was excluded because it did not report their data to NCANDS, and Pennsylvania was excluded because it did not report the race or ethnicity of the children. Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model
ARTICLE IN PRESS
CHIABU-3005; No. of Pages 13
K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
5
Table 1 Rural-Urban Continuum Codes (RUCC). Code
Description
Metropolitan counties 1 2 3
Counties in metro areas of 1 million population or more Counties in metro areas of 250,000 to 1 million population Counties in metro areas of fewer than 250,000 population
Non-metropolitan counties 4 5 6 7 8 9
Urban population of 20,000 or more, adjacent to a metro area Urban population of 20,000 or more, not adjacent to a metro area Urban population of 2,500 to 19,999, adjacent to a metro area Urban population of 2,500 to 19,999, not adjacent to a metro area Completely rural or less than 2,500 urban population, adjacent to a metro area Completely rural or less than 2,500 urban population, not adjacent to a metro area
The county maltreatment rate per 1,000 children was calculated by dividing the count of reported children by the 2012 child population produced by the National Center for Health Statistics (NCHS, 2014) using Census data. Second, the county family poverty rate (households with related children under age 18) per 100 families was calculated using the US Census American Community Survey (ACS) 2008–2012 5-year estimates. Third, county child population density (log transformed) was obtained from the 2010 US Census. We also examined geographic type by relying on the 2013 Rural-Urban Continuum Codes (RUCC). The United States Department of Agriculture Economic Research Service develops these codes. The RUCC classification scheme categorizes each county into 9 groups based on population size, degree of urbanization, and adjacency to metro areas based on the 2010 Census and the 2006–2010 ACS. Table 1 provides the description of each RUCC. For our key geographic measure, we rely on RUCC in our exploratory analyses and then population density in multivariate analyses because population density is a ratio-level, continuous variable.
Data Analysis County-level data from NCANDS, Census, and RUCC were linked using the county FIPS code. Maltreatment rates and poverty rates were calculated for whites, blacks, and Hispanics. Black–white and Hispanic–white disparity ratios (DRs) were calculated by dividing the minority (black or Hispanic) maltreatment or poverty rate by the corresponding white rate. Using this approach to calculate DRs, the DR equals 1 when the minority rate is equal to the white rate. A higher DR corresponds to greater ratio of the minority rate to the white rate in the same county. For example, a DR equal to 2 indicates that the minority rate is twice that of the white rate. A DR below 1 indicates that the minority rate is lower than the white rate. Using county as the geographic unit of aggregation has advantages including the availability of data, reduction of concerns about confidentiality, and reliability of measures (given that maltreatment is a relatively rare event, smaller areas may not be stable in terms of the rates of such events; Aron et al., 2010). However, there are also drawbacks to examining counties. Many counties have small population sizes overall or for specific race/ethnic groups yielding outliers. Comparing rates for such small populations yields unstable results, as a single maltreatment report can produce large changes in the rate. We addressed small county population sizes with two approaches. First, we pooled counties (first nationally and then within each state) based on RUCC and calculated rates for counties of the same RUCC. This approach pools rural areas with smaller populations into one group to calculate a rate for all similar rural areas within the same state. Pooling yielded 348 pooled county groups across the 48 states (not all states have counties representing all nine RUCC groups). Second, for the countylevel analysis, we flagged all counties with
ARTICLE IN PRESS Child Abuse & Neglect xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Child Abuse & Neglect
Research article
Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density Kathryn Maguire-Jack a,∗ , Paul Lanier b , Michelle Johnson-Motoyama c , Hannah Welch b , Michael Dineen d a b c d
The Ohio State University College of Social Work, 325B Stillman Hall, 1947 N College Road, Columbus, OH 43210, USA University of North Carolina Chapel Hill, 325 Pittsboro Street, CB #3550, Chapel Hill, NC 27599, USA University of Kansas School of Social Welfare, 311 Twente Hall, 1545 Lilac Lane, Lawrence, KS 66045, USA National Data Archive for Child Abuse and Neglect (NDACAN), Cornell University, Ithaca, NY 14853, USA
a r t i c l e
i n f o
Article history: Received 23 March 2015 Received in revised form 25 May 2015 Accepted 27 May 2015 Available online xxx Keywords: Child maltreatment Spatial Hierarchical linear modeling Geographic variation Poverty Population density
a b s t r a c t There are documented disparities in the rates at which black children come into contact with the child welfare system in the United States compared to white children. A great deal of research has proliferated aimed at understanding whether systematic biases or differential rates of risk among different groups drive these disparities (Drake et al., 2011). In the current study, county rates of maltreatment disparity are compared across the United States and examined in relation to rates of poverty disparity as well as population density. Specifically, using hierarchical linear modeling with a spatially lagged dependent variable, the current study examined data from the National Child Abuse and Neglect Data System (NCANDS) and found that poverty disparities were associated with rates of maltreatment disparities, and densely populated metropolitan counties tended to have the greatest levels of maltreatment disparity for both black and Hispanic children. A significant curvilinear relationship was also observed between these variables, such that in addition to the most densely populated counties, the most sparsely populated counties also tended to have higher rates of maltreatment disparity for black and Hispanic children. © 2015 Elsevier Ltd. All rights reserved.
Introduction The United States child welfare system (CWS) has a history of differential treatment for children of different races. In the past, minority children were excluded from the CWS entirely, with black children initially excluded from the developing foster care system (Roberts, 2002). However, black and Hispanic children are currently represented in the CWS at rates that are different than their representation in the general population. In 2013, black children made up 15% and Hispanic children made up 23% of the total child population, but they were represented among child maltreatment victims at proportions of 21% and 22%, respectively (United States Census Bureau, 2014a; United States Department of Health and Human Services, 2015). Additionally, a recent study of the cumulative prevalence of confirmed child maltreatment by age 18 years estimates that black children have a cumulative prevalence of 20.9%, while the rates for Hispanic and white children are 13.0% and 10.7%, respectively (Wildeman et al., 2014).
∗ Corresponding author. http://dx.doi.org/10.1016/j.chiabu.2015.05.020 0145-2134/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
2
An abundance of research has documented associations between child maltreatment and negative long-term outcomes that last into adulthood. Children who are maltreated are more likely to struggle with behavioral problems (Felitti et al., 1998; Repetti, Taylor, & Seeman, 2002), mental health conditions (Holmes & Sammel, 2002), increased risk for criminal behavior and violent behavior (Fang & Corso, 2007; Widom & Maxfield, 2001), lower health-related quality of life (Corso, Edwards, Fang, & Mercy, 2008; Lanier, Kohl, Raghavan, & Auslander, 2015) and lower long-term economic well-being (Currie & Widom, 2010). Given the disproportional representation of black and Hispanic children among child maltreatment victims and the significant sequelae for victims, it is essential that we understand why racial disparities occur. If overrepresented children are not involved in this system as a result of need, but rather, systematic bias on the behalf of CPS workers, changes must be made immediately to reverse this trend. If, however, this phenomenon results from disproportionate risk of child maltreatment due to other inequalities within our society, then the problem is not that ethnic minority children are overserved by CPS, but rather, that they are maltreated at a higher rate than their non-minority peers. Under the latter scenario, intervention needs to occur prior to CPS involvement. In addition to variation by child/race ethnicity, child maltreatment rates vary geographically (Ernst, 2001; Fryer & Miyoshi, 1995; Paulsen, 2003; Vinson & Baldry, 1999; Zielinski & Bradshaw, 2006). At the county level, researchers have found associations between a number of poverty-related (Albert & Barth, 1996; Freisthler & Weiss, 2008; Hon-Yei, 1989; Spearly & Lauderdale, 1983) and substance use-related (Albert & Barth, 1996; Freisthler & Weiss, 2008) factors and child maltreatment rates, with important differences when comparing urban, suburban, and rural areas (Albert & Barth, 1996). These areas differ in terms of service accessibility and availability, social connections, and poverty rates (Slovak & Carlson, 2009), which may have important implications for rates of racial disparity in child maltreatment. To the extent that ethnic minority families are concentrated in counties with higher levels of poverty and lower access to resources that address these challenges, they may be at an increased risk for child maltreatment; thus contributing to the racial disparity in child maltreatment. For example, in a study of aboriginal children in Canada, it was found that race and ethnicity of individual families was not driving differential placement rates, but rather, certain characteristics of agencies in areas with high concentrations of aboriginal children, which the authors attributed to a lack of resources for aboriginal children (Fluke, Chabot, Fallon, MacLaurin, & Blackstock, 2010). In the U.S., Hispanic families are significantly more likely to live in highly segregated communities that have higher rates of social isolation and greater resource deprivation (Hipp & Yates, 2011; Lee, 2000; Parker & McCall, 1999). Further, neighborhoods that are highly segregated with a high concentration of black residents are less likely to have access to basic needs assistance, mental health and substance abuse providers, and employment services (Allard, 2009). Therefore, the disproportional representation of ethnic minority children in the U.S. CWS may be partly explained by disproportional residence in high poverty geographic areas. Theories of Disparity We use the term “disparity ratio (DR)” to refer to the ratio of the rate of a phenomenon in one racial group compared to the rate in another racial group and “disparity” to refer to the extent to which the ratio is not equal to 1 (Myers, 2010). It should be noted that while some researchers have used the term “disparity” to refer explicitly to unequal treatment of a minority group (Hill, 2006), we are not referring to that definition of the term. We use “disparity” to refer simply to the existence of differences in rates of phenomena, and do not assume that the cause of such difference is due to unequal treatment. Black–White Disparity Among researchers and advocates, there have been three major arguments to explain the overrepresentation of black children in the CWS. All three of these theories are based on the hypothesis that the long and complicated history of racial discrimination in the United States has had profound and lasting impacts on families, which contribute to disparity. The key difference between these theories lies in the level at which discrimination occurs and the mechanism through which it contributes to disparity. The first hypothesis is that racial bias at the individual level causes caseworkers to interpret observations differently for black children compared to white children (Chand, 2000; Webb, Maddocks, & Bongilli, 2002). In this scenario, caseworkers are thought to be more likely to substantiate maltreatment allegations and remove a child into out-of-home care when the child is black compared to a white child. A second theory, proposes that institutional racism has created policies and procedures for mandated reporters, social service caseworkers, and family court judges that are inherently biased against black children and families (Ards, Myers, Malkis, Sugrue, & Zhou, 2003; Hill, 2004). The third and final explanation is that current and historic discrimination has put black families at a significant disadvantage economically, which increases their risks for maltreatment (e.g. poverty, single parenthood, substance use), and as a result, black families have higher rates of maltreatment (Bartholet, Wulczyn, Barth, & Lederman, 2011; Drake, Jolley, & Lanier, 2011; Font, Berger, & Slack, 2012; Lanier, Maguire-Jack, Walsh, Drake, & Hubel, 2014). Though official statistics show that black residents of the United States are much more likely to live below the poverty level (approximately 27% of blacks compared to 10% of whites; U.S. Census Bureau, 2014b), such numbers significantly understate the level of socioeconomic disadvantage black families experience. Black children are roughly three times more likely to be poor, but are up to 14 times more likely to live in high childhood poverty neighborhoods compared to white children (Drake & Rand, 2009). Further, blacks have roughly 1/10 the assets of whites in the United States (Kochhat, 2015). Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
3
Despite higher rates of referrals overall regarding black children, prior research has shown that white children in higher poverty areas are more likely than black children to be the subject of a child maltreatment report (Drake, Lee, & JonsonReid, 2009). Drake and Rand (2009) proposed a complementary theory to the theories of racial disparity, referred to as the theory of differential assortment. This theory posits that given higher rates of wealth and assets and relatively unlimited mobility among white families who have not experienced the same historic forms of discrimination faced by black families (e.g. redlining), whites are much less likely to live in areas of high rates of poverty. When white families do live in highly impoverished areas, they are likely to be there due to a multitude of risk factors (e.g., substance abuse, mental health issues) when compared to black families, who are more likely to live in such areas as a result of racial segregation, regardless of individual risks. Therefore, it is expected that white families in poverty-stricken communities should have higher rates of child maltreatment reports compared to black families in similar communities, by virtue of differential levels of risk.
Hispanic–White Disparity Despite local variation in state and county estimates of disparity among Hispanics yielding observations of both overrepresentation and underrepresentation (Ards, Myers, Malkis, Sugrue, & Zhou, 2003; Church, Gross, & Baldwin, 2005; Dettlaff et al., 2011; Needell et al., 2012), theories regarding the disproportional representation of Hispanic children in the CWS have predominately focused on explaining the underrepresentation that has been observed in child maltreatment at the national level (Sedlak, 2010; United States Department of Health and Human Services, 2015). These theories have focused on trends that have been described as the “Hispanic health paradox,” as well as the differential surveillance of Hispanic families, particularly among families with immigrant members. Health paradoxes refer to patterns of morbidity and/or mortality for a particular group that is at odds with what would be expected given the group’s socioeconomic profile (Acevedo-Garcia & Bates, 2008). Such trends have been observed for certain infant and adult health outcomes, whereby Hispanics, particularly immigrants, often demonstrate better health when compared to their white and non-immigrant counterparts of similar socioeconomic position (Franzini, Ribble, & Keddie, 2001; Hayes-Bautista, 2002; Jasso, Massey, Rosenzweig, & Smith, 2004; Palloni & Arias, 2004; Palloni & Morenoff, 2001). The presence of health advantages among Hispanics and immigrants, despite socioeconomic disadvantage, has been attributed widely to protective social and cultural dynamics (e.g., familism, religiosity, salutary health behavior). This social-cultural explanation is often presented in association with an acculturation hypothesis that predicts the deterioration of health among Hispanics over time and across generations in the U.S. (i.e., health assimilation) (see Acevedo-Garcia & Bates, 2008). Another explanation for the health advantages among immigrants is referred to as the “healthy immigrant effect,” which posits that those who are able to migrate have certain health advantages that provide them with physical and financial abilities to migrate (McDonald & Kennedy, 2004). Recent studies arising from population-level linkages of child welfare and vital statistics data in California have revealed disparities in CWS involvement that are consistent with the Hispanic health paradox, whereby Hispanic children of foreignborn and U.S.-born Hispanics are less likely to be involved with the CWS when compared to their white counterparts after adjusting for socioeconomic position (Putnam-Hornstein, Needell, King, & Johnson-Motoyama, 2013). Studies have also identified the significant challenges that Hispanic immigrant caregivers in lower-income households face in meeting their children’s basic needs for food, medical care, and safe supervision (Johnson-Motoyama et al., 2014). Despite these risks, infants of Hispanic foreign-born mothers are consistently less likely to become involved with the CWS when compared to infants of Hispanic U.S.-born mothers across ethnic subgroups (e.g., Mexican, Central American) (Johnson-Motoyama et al., 2014). A second theory regarding the underrepresentation of Hispanics in national estimates of child maltreatment victimization posits underreporting as a result of differential surveillance (Putnam-Hornstein et al., 2013). The differential surveillance of maltreatment among Hispanic families has been attributed to differential exposure to mandated reporters in social, health, and child care settings due to lower utilization of these services (Flores et al., 2002; Karoly & Gonzalez, 2011), particularly among undocumented families whose access to health and social assistance are typically restricted by policies and practices at the federal level and in many states and localities. Among legal permanent residents, access to health and social programs typically require waiting periods, which may influence exposure to mandated reporters. Moreover, national and local evaluations suggest that the eligibility restrictions embedded in anti-immigrant policy initiatives successfully reduce the participation of immigrant families in public programs, regardless of legal status (Borjas, 2011; Fix & Passel, 2002; Zimmerman & Fix, 1998), as policies that negatively impact immigrant caregivers typically have downward effects on citizen children’s participation in public programs (Fix & Zimmerman, 2001; Seiber, 2013). Such trends may reduce the number and types of contacts that Hispanic children have with mandated reporters. Underrepresentation may also arise from the differential mobility of immigrants. For example, many Mexican and Central American immigrants cross the U.S.–Mexico border throughout the year in seasonal patterns that correspond with U.S. labor demand (Dockterman, 2011; Passel & Cohn, 2009). Hispanic children, particularly those with foreign-born mothers, may be at lower risk for CWS involvement when compared to whites and children of native-born Hispanic mothers due to migration patterns that limit exposure to mandated reporters. Finally, some researchers suggest that paradoxical patterns might be due to inconsistent ways of defining Hispanic identity (Franzini, Riblle, & Keddie, 2001; Jasso et al., 2004; Palloni & Morenoff, 2001). Given the limited ability of most studies to Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model CHIABU-3005; No. of Pages 13
ARTICLE IN PRESS K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
4
account for all of these possible explanations, the question of whether a Hispanic health paradox exists with regard to child maltreatment is far from settled. Background Drake et al. (2011) sought to test some of these explanations for disparity by comparing national rates of child maltreatment reports from the National Child Abuse and Neglect Data System (NCANDS). They compared disparity in child maltreatment to disparity in poverty as a risk factor for maltreatment, as well as disparity in a number of other negative outcomes for children that were deemed to be less susceptible to bias including infant mortality, low birth weight, premature birth, infant accidents, and infant homicides. The authors found that the black–white disparity in child maltreatment was very similar to the rates of disparity in poverty as well as all of the other negative outcomes for children. As a result, they concluded that child maltreatment disparity was likely the result of increased risk for maltreatment among black families, and not a function of reporter bias. Although the Hispanic–white disparity ratio for poverty was similar to that for black children compared with white children, data suggested that the presence of protective moderating factors offset the relationship between poverty and poor outcomes for Hispanic children, lending support to the notion of a Hispanic health paradox in child maltreatment (Drake et al., 2011). In an extension of this work, Lanier et al. (2014) conducted a state by state comparison of child maltreatment risk measures (e.g., rates of single mothers, rates of births to teen mothers) and found child maltreatment disparity to be highest in states with high rates of poverty disparity, providing additional support for the findings by Drake et al. (2011). On average, the level of maltreatment disparity for Hispanics was lower than that of blacks; however, the relationships of maltreatment to poverty remained consistent across the two groups to suggest the presence of protective factors that potentially mitigate the risk of maltreatment among Hispanic children (Lanier et al., 2014). Given the variation in maltreatment disparity that has been observed at the local level, particularly among Hispanic children, the current study delves deeper into these relationships by examining the trends of child maltreatment and poverty disparity at the county level. Specifically, we examine three questions: (1) Is there a relationship between poverty and maltreatment such that counties with higher disparity in poverty also have higher maltreatment disparity?; (2) Is there a relationship between maltreatment disparity and county population geography (i.e., rural vs. urban)?; and (3) Do these relationships differ for black–white disparity when compared to Hispanic–white disparity? A related study examined the relationship between county-level poverty and county-level racial disparities in foster care placement between black and white children (Wulczyn, Gibbons, Snowden, & Lery, 2013). In that study, the authors found that areas with higher levels of white poverty had a smaller gap between black and white rates of foster care placement; and that black poverty rates largely did not explain the difference in the rates. Our study builds on this prior work in three important ways (1) by focusing on an earlier step in CPS decision-making, that is, the rate of child maltreatment reports rather than foster care placements; (2) by examining disparities in poverty rates by race, rather than examining the rate for each race separately; and (3) by examining these relationships in counties in 48 states as well as Washington DC rather than 16 states. In doing so, we seek to advance knowledge about how to effectively reduce racial disparities in maltreatment. Method This study examined the association between three county-level indicators for white, black, and Hispanic children: maltreatment, poverty, and geography. We used two approaches to explore this relationship. First, we used mapping techniques to visually observe county-level variation in the indicators across the United States. Second, we used multivariate regression analyses to statistically test for observed relationships. Non-Hispanic white, non-Hispanic black, and Hispanic groups (without reference to race) were included because data for these groups were available in most states for the child indicators used in this study and were of key interest to expand on prior research, theory, and practice. Area Definitions County was chosen as the geographic unit of interest in this study. Prior work from Lanier et al. (2014) suggested statelevel variation in maltreatment DR, which prompted our interest in examining whether the large geographic unit of a state was masking important variation within. We selected county as the unit of analysis because it was the smallest level of geography available within the child maltreatment dataset. Data Sources and Variables Three data sources were linked at the county level using the Federal Information Processing Standard (FIPS), a unique spatial code. First, the unduplicated count of children who were the subject of a child maltreatment report was obtained from the 2011 NCANDS Child File. NCANDS data contain all reports of maltreatment to CPS for all states (and Washington, DC) except Oregon and Pennsylvania, which were excluded from the analysis. Oregon was excluded because it did not report their data to NCANDS, and Pennsylvania was excluded because it did not report the race or ethnicity of the children. Please cite this article in press as: Maguire-Jack, K., et al. Geographic variation in racial disparities in child maltreatment: The influence of county poverty and population density. Child Abuse & Neglect (2015), http://dx.doi.org/10.1016/j.chiabu.2015.05.020
G Model
ARTICLE IN PRESS
CHIABU-3005; No. of Pages 13
K. Maguire-Jack et al. / Child Abuse & Neglect xxx (2015) xxx–xxx
5
Table 1 Rural-Urban Continuum Codes (RUCC). Code
Description
Metropolitan counties 1 2 3
Counties in metro areas of 1 million population or more Counties in metro areas of 250,000 to 1 million population Counties in metro areas of fewer than 250,000 population
Non-metropolitan counties 4 5 6 7 8 9
Urban population of 20,000 or more, adjacent to a metro area Urban population of 20,000 or more, not adjacent to a metro area Urban population of 2,500 to 19,999, adjacent to a metro area Urban population of 2,500 to 19,999, not adjacent to a metro area Completely rural or less than 2,500 urban population, adjacent to a metro area Completely rural or less than 2,500 urban population, not adjacent to a metro area
The county maltreatment rate per 1,000 children was calculated by dividing the count of reported children by the 2012 child population produced by the National Center for Health Statistics (NCHS, 2014) using Census data. Second, the county family poverty rate (households with related children under age 18) per 100 families was calculated using the US Census American Community Survey (ACS) 2008–2012 5-year estimates. Third, county child population density (log transformed) was obtained from the 2010 US Census. We also examined geographic type by relying on the 2013 Rural-Urban Continuum Codes (RUCC). The United States Department of Agriculture Economic Research Service develops these codes. The RUCC classification scheme categorizes each county into 9 groups based on population size, degree of urbanization, and adjacency to metro areas based on the 2010 Census and the 2006–2010 ACS. Table 1 provides the description of each RUCC. For our key geographic measure, we rely on RUCC in our exploratory analyses and then population density in multivariate analyses because population density is a ratio-level, continuous variable.
Data Analysis County-level data from NCANDS, Census, and RUCC were linked using the county FIPS code. Maltreatment rates and poverty rates were calculated for whites, blacks, and Hispanics. Black–white and Hispanic–white disparity ratios (DRs) were calculated by dividing the minority (black or Hispanic) maltreatment or poverty rate by the corresponding white rate. Using this approach to calculate DRs, the DR equals 1 when the minority rate is equal to the white rate. A higher DR corresponds to greater ratio of the minority rate to the white rate in the same county. For example, a DR equal to 2 indicates that the minority rate is twice that of the white rate. A DR below 1 indicates that the minority rate is lower than the white rate. Using county as the geographic unit of aggregation has advantages including the availability of data, reduction of concerns about confidentiality, and reliability of measures (given that maltreatment is a relatively rare event, smaller areas may not be stable in terms of the rates of such events; Aron et al., 2010). However, there are also drawbacks to examining counties. Many counties have small population sizes overall or for specific race/ethnic groups yielding outliers. Comparing rates for such small populations yields unstable results, as a single maltreatment report can produce large changes in the rate. We addressed small county population sizes with two approaches. First, we pooled counties (first nationally and then within each state) based on RUCC and calculated rates for counties of the same RUCC. This approach pools rural areas with smaller populations into one group to calculate a rate for all similar rural areas within the same state. Pooling yielded 348 pooled county groups across the 48 states (not all states have counties representing all nine RUCC groups). Second, for the countylevel analysis, we flagged all counties with
