A Multilevel Analysis of Individual and Neighborhood Level Predictors of Child Maltreatment

Background & Purpose: Child maltreatment is an enduring problem in the United States with well-documented negative effects on children.  Though the literature suggests a consistent relationship between neighborhood impoverishment and rates of child abuse and neglect, methodological concerns may limit the validity of these findings. The present study uses hierarchical linear modeling to address these methodological concerns and to examine the impact of individual level poverty, concentrated neighborhood disadvantage, and other neighborhood characteristics on child maltreatment outcomes. 

The purposes of this study were to 1) test the relationship between individual level poverty and length of a child’s active Children, Youth, and Families (CYF) case, 2) test the relationship between contextual (neighborhood) level poverty and the length of a child’s active CYF case, 3) examine the extent to which neighborhood level poverty and individual level poverty contribute to length of a child’s active CYF case, controlling for neighborhood and child level characteristics.

Methods: The sample in this study was drawn from a hierarchical dataset comprised of data obtained from the Allegheny County Department of Human Services (DHS) and Census data from the American Community Survey 5 year estimates (2005-2009). Using Geographic Information Systems (GIS) all children were joined to a census tract based on spatial data included in the DHS database.  The resulting sample included 3,703 children in 125 census tracts in the City of Pittsburgh.  The dependent variable was the length of a child’s active CYF case. Independent variables at the child level include poverty, race, and age. Independent variables at the contextual level included an index of concentrated neighborhood disadvantage, percentage of vacant properties, and percentage of owner occupied housing units.  Models were estimated using hierarchical linear modeling in HLM 7.0 software.

Results: Findings demonstrated a significant positive relationship between child poverty and length of an active child abuse and neglect case (p<.001). Contrary to prior research, the findings of this study suggested that after controlling for child level poverty, concentrated neighborhood disadvantage and associated neighborhood variables became non-significant. Only about 3% of the variance in child maltreatment outcomes was attributable to neighborhood (p<.001).  However, in neighborhoods with lower percentages of vacant properties, the impact of concentrated neighborhood disadvantage exerted a stronger influence (p<.001).  This suppressive effect lends support to further exploration of physical neighborhood characteristics as moderators in the relationship between child maltreatment and poverty.

Implications:  This study sought to address methodological concerns in the current literature around individual and neighborhood effects on child maltreatment. Consistent with prior research, this analysis suggests that child poverty is a significant predictor of maltreatment. Stresses associated with living in poverty may have an effect on parents that leads to maltreatment or may lead, due to a lack of resources, to neglect. Though the effect sizes were small, there is some connection between neighborhood of residence and the length of a child’s involvement in CYF services.  This points to the need to further explore structural factors related to child maltreatment in research, policy, and practice.