Evidence implicates schools districts and counties as potentially important units of analysis for understanding linkages between child maltreatment indicators and school performance. There is area-level variation in child maltreatment rates (Coulton et al., 2008) and both maltreated and children in out-of-home care are distributed non-randomly across schools and districts (Stone, 2007). Critical densities of foster children, net of measures of neighborhood socio-demographic and social capital may interfere with functional school-level organizational process and achievement trajectories, leading to the hypothesis that aggregate child well-being characteristics represent key “contextual” effects on schools'/districts' performance.
Methods: This study draws on four longitudinal data sources to estimate the association between county-level maltreatment rates on school district achievement in California. The first includes county-level estimates of substantiated and inconclusive referral rates for calendar years 2001-2007. County level population estimates of child health were generated from the California Health Interview Surveys (CHIS, 2001, 2003, 2005, 2007). Population-level data on school districts were obtained from California Department of Education archives for the 2001-2002 through 2008-2009 school years. Finally, county demographic characteristics are derived from census and vital statistics. The study sample encompasses 934 districts (in 53 out of 58 total counties) and represents 96% of districts with achievement data and 94% of districts in California.
The outcome is a dichotomous variable indicating whether, in the time period between 2001-2002 and 2008-2009, a district had either consistently failed to meet its adequate yearly progress goals or had changed status from meeting to not meeting these goals. The key independent variable is child welfare system referral rates. A rich set of district- and county-level characteristics (e.g. district demographic, structural, and achievement characteristics; county socio-demographic and child health characteristics) serve as controls.
Because there is great potential for biased estimates of associations between maltreatment rates and school districts, the longitudinal and time-lagged structure of these data is used to implement fixed effect models to control for the influence of unobservable, time-invariant district and county characteristics. Standard errors in regression models are adjusted for the clustering of districts within counties.
Results: County change in child and adolescent maltreatment rates increased the likelihood that a school district failed to meet “adequate yearly progress” goals over the time period under investigation. That is, districts in counties with fewer maltreated children and adolescents, over and above demographic, child health, and district demographic and performance trajectories, were less likely to sustain or improve academic performance over time.
Conclusions: This study provides empirical support for the contention that area variation in child and adolescent maltreatment exerts a contextual effect on school district performance and suggests that aggregate levels of child well-being may itself function as a protective setting-level factor.