School District and Individual Predictors of Foster Care Students' Academic Proficiency

Background and Purpose:

Research consistently demonstrates that children and youth involved in the child welfare system experience poorer school outcomes than non-child welfare involved peers. These outcomes include lower math, reading, and achievement test scores, lower grades, and higher rates of grade retention, special education service utilization, school behavior problems, school suspension, and high school dropout. To understand which factors impact foster children’s academic outcomes, above and beyond individual characteristics, this study incorporates relevant school district-level covariates that have not been commonly accounted for in analysis of foster children’s school achievement.

The study builds on the strengths and limitations of previous work in answering the research question: “To what extent do individual and school district factors influence foster care students’ academic proficiency test scores?” By focusing on higher-level influences on student outcomes, this study can help inform school district-level policies and practices to improve academic outcomes for youth in the child welfare system. Further, in service of the long-term goal of clarifying the causal pathways from foster care involvement to education outcomes, this study attends to the intermediate goal of identifying and specifying a comprehensive set of predictors and confounders of these outcomes.  

 

Methods:

This study employed multilevel linear modeling (MLM) with individual and school district-level data taken from state administrative data sources in a large, northeastern state. A sample of foster children (n=2037) with records from child welfare, education, and court datasets was utilized in the study. Students’ English Language Arts (ELA) and Math scaled academic proficiency test scores were modeled as the outcomes of interest, and individual-level demographic, child welfare, and school experience variables as well as district-level variables were included in the analysis. Models were fit with random intercepts at the district level, assuming fixed slopes for the covariates.

 

Results:

Analyses indicated that district differences account for 4.28% of variation in foster children’s ELA test scores and 2.12% of variation in math test scores. These small but important measures of district-level variance were robust to the introduction of individual and district-level predictors and confounders. In the full models, significant individual-level predictors of the test score outcomes included special education involvement, being overage for grade, and specific child welfare placement types (all associated with lower scores). Significant district-level predictors included proportion of low-SES students (lower scores) and spending per pupil (higher scores).

 

Conclusions and Implications:

This study contributes meaningfully to the child welfare literature by identifying variation in foster children’s academic proficiency scores at the school district level. The results of this study highlight important, measureable, and policy-relevant district-level predictors of test scores, specifically expenditures per pupil, as well as quantifying the magnitude of district-level sources of test score variation. The study results have policy and research implications, suggesting district-level per-pupil spending may be a lever to positively influence foster children’s academic outcomes and underscoring the importance of examining other district-level predictors of foster children’s school outcomes.