Dimensions of Structural Disadvantage: A Latent Class Analysis of a Neighborhood Measure in Child Welfare Data

Background/Purpose: It is well-documented that child abuse and neglect reports come primarily from more disadvantaged neighborhoods with high unemployment, economic deprivation, residential turnover, and poverty. However, less is known about how neighborhood dimensions might predict the type of child welfare system involvement, which can be used to inform child maltreatment prevention efforts and help build community interventions. Studies that focus on subjective measures of the neighborhood, such as asking parent’s perceptions of their place of residence, are also less frequent in this literature. In this paper, I conducted an ex post factolatent class analysis of caregiver neighborhood perceptions after an abuse or neglect investigation from a nationally representative sample of children investigated by child protective services.

Methods: This paper used data from the nationally representative, longitudinal National Survey of Child and Adolescent Well-Being – II (NSCAW-II). I drew on the abridged version of the Community Environment Scale (CES), where caregivers assessed risk and protective factors in their neighborhood, such as self-perceived crime, safety, neighborliness, and information about other families with children. I completed latent class analysis (LCA) with the CES, which the child’s caregiver completed at baseline. LCA was ideal for this study to identify meaningful groups of respondents using neighborhood risk factors facing families involved with the child welfare system. LCA groups respondents based on similar patterns, which is ideal when considering contextual risk factors among the child welfare population.

Results: In order to determine the best fitting model, I calculated a one-class solution to a ten-class solution. The three-class model was chosen based on model interpretability and parsimony. The three classes are labeled: high social order, medium social capital (35.0% of respondents); high social order, low social capital (47.7% of respondents); and low social order, low social capital (17.3% of respondents). Multinomial logistic regression analysis partially validated findings from the LCA. Black and Hispanic children and families with lower incomes, low social support, reports of physical abuse, and prior reports to the child welfare are more likely to reside in neighborhoods with the highest risk factors – those with low social order (higher crime) and lower social capital (fewer neighbors that help each other).

Conclusions/Implications: Simply understanding neighborhoods as “high” versus “low” risk may not fully illuminate contextual factors to develop neighborhood-based child welfare interventions. In fact, these results point to a cluster of families who reside in neighborhoods that might appear low risk based on social order, but have a lack of social capital, which is an important measure for understanding family functioning and child development. In the future, comparisons can be made using latent classes to further identify differences in community factors for at-risk populations and how residential context might contribute to child maltreatment and aid in child welfare interventions. State and local agencies may be able to develop a community approach to prevent child maltreatment before it starts. Similar risk factors might be contributing to maltreatment for multiple families in a neighborhood, which might make a community approach both cost-effective and successful.