Neighborhood Consistency in Levels of Unusual Risk for and Protection from Child Maltreatment Referrals

Background and Purpose: Individual family-level variables strongly predict future likelihood of involvement with child welfare services. For example, parental age, education, income, family composition, and number of children in the home substantially aid in understanding risk for child welfare involvement. Child maltreatment, however, can also be viewed as a highly place-based phenomenon that affects neighborhoods and sub-neighborhood level areas. As demonstrated in diverse regions, child maltreatment referrals tend to cluster geographically. Such concentration occurs due to social and institutional characteristics of neighborhoods, as well as social sorting forces that bound family living choices. As such, the aims of this study are to: 1) model predictors of child maltreatment referrals at the neighborhood (census tract) level in two large California counties, 2) estimate the frequency of neighborhood areas with unusual risk for or protection from child maltreatment referrals after controlling for population characteristics, and 3) examine whether neighborhoods have year-over-year consistency in unusual levels of risk and protection.

Methods: Child maltreatment referral data aggregated at the census tract level from calendar years 2012 and 2013 were combined with American Community Survey data from the US Census bureau. Linear regression models estimated referral rates per 100 children in each census tract separately in San Diego and Los Angeles counties. Unusual areas of risk and protection were identified, defined as census tracts with maltreatment referral rates either 3 or more referrals per 100 children above or below predicted levels based on local population characteristics. Residual referral rates in 2013 were compared with referral rates in previous years to understand year-over-year consistency in neighborhood areas appearing to have increased risk or protection.

Results: In San Diego and Los Angeles counties, approximately 5 referrals occurred per 100 children in the average census tract. Population models of referral rates explained a large percentage (45%-55%) of referral rate variance each year at the neighborhood level, with the same core variables having greatest predictive influence in both counties (education, poverty, single parenthood, limited labor force involvement, and housing transiency; all significant). 10-15% of neighborhoods had unusually high or low maltreatment referral rates in both counties each year. Among census tracts with unusual levels of risk in 2012 (average of 5.8 referrals more than predicted), they continued to have 3.4 referrals more than predicted in 2013. Among census tracts with unusual levels of protection in 2012 (average of 4.0 fewer referrals than expected), they continued to have 3.1 referrals fewer than predicted in 2013.

Conclusions:  Areas of risk for and protection from child maltreatment referrals have strong year-over-year consistency. This occurs after including aggregate population predictors derived from individual family risk factors, suggesting that characteristics of environments in which individuals live may strongly shape maltreatment risk. We will discuss our methods for identifying neighborhoods with unusually high and low maltreatment referral rates and how cutoffs were selected. We will further discuss potential explanations from our ongoing research regarding community-level characteristics that may explain such consistency and possible implications for community-focused prevention strategies.