Methods. County-level data on child maltreatment reports and victimization from 2015 were merged with data from the U.S. Census and the National Child Abuse and Neglect Data System for all U.S. counties with a population of at least 1,000 (n = 2,966). Key independent variables at the county level included the child poverty rate, percentage rural, and percentages African-American, white, and Latinx. State-level variables included whether the administrative structure was state or county-based, whether alternative response protocols were in place, the child welfare worker/child rate, and an indicator of Medicaid accessibility. Bivariate correlations and HLM multi-level regression models (with counties nested in states) were conducted to assess associations between administrative structure and the county-level report and victimization rates.
Results. The county-level investigated report rate ranged from less than 1/1,000 children to over 120/1,000 children, and the victimization rate ranged from less than 1/1,000 to 42/1,000. At a bivariate level, the report rate (39.3 state vs. 27.7 county-administered; t = 16.78, p<.001) and victimization rate (8.2 state vs. 5.3 county-administered; t = 13.14, p<.01) were significantly higher in counties with state-administered systems. In multi-level models accounting for child poverty, demographic, and administrative characteristics, maltreatment report and victimization rates remained significantly higher in states with state-administered child welfare systems, supporting both parts of the hypothesis. Multi-level models also confirmed that substantial variation in both reports (icc = .46) and victimization (icc = .48) is at the state level.
Implications. To our knowledge, this is the first study to assess the association of child welfare administrative structures on county-level maltreatment report and victimization rates for all U.S. counties. The “three pillars” of institutional theory (Scott, 1995) illuminate how regulatory, normative, and cultural/cognitive processes can affect organizational processes. These findings suggest that child welfare practitioners and other decisionmakers in county-administered child welfare systems may feel different pressures, rely on different norms, and hold shared beliefs that affect practice in ways that result in fewer investigated reports and fewer substantiated reports. Perhaps decentralized child welfare systems empower frontline decisionmakers to apply a shared logic of action, which further shapes perception and discretion (Sosin, 2009) to more heavily weigh family protective factors, and draw more on an institutional network of community resources.