Method: This study used administrative data from ~12,000 students enrolled in 121 Baltimore City Public Schools from 2007-2013 (85% African American; 83% low-income). School readiness assessments were collected in fall of kindergarten using the Maryland Model for School Readiness (MMSR), a portfolio-based assessment measuring 7 domains: social-behavioral skills, language/literacy, mathematics, social studies, scientific thinking, physical development, and arts. Student demographics (kindergarten entry age, gender, race/ethnicity, low-income status), and whether students were chronically absent from kindergarten, English Language Learners in kindergarten, and attended public PreK were included as covariates. Multi-level modeling, accounting for nesting within schools where students took the MMSR, was used to examine associations between social-behavioral readiness in kindergarten, kindergarten entry age and public Prek participation, and three outcomes through 3rd grade: being suspended/expelled, receiving special education services, being retained.
Results: Initial multi-level models indicated that after controlling for all other MMSR domains, students not identified as socially-behaviorally ready for kindergarten were more likely to experience all three outcomes. Analyses to understand whether kindergarten entry age and participation in public preK moderate these relations are in process. Key challenges encountered in using administrative data included: 1) data/measure quality issues (e.g., social-behavioral readiness measure failed to tap all aspects of construct), and 2) how best to handle missing data due to students moving out of and back in to the district over the course of the study.
Conclusions: Findings and challenges have implications for use of big administrative datasets to inform educational policies and practices.