Our team's recent work applying multi-level modeling to early childhood education outcomes illustrates how evaluation can move beyond surface-level findings to uncover how programs function across interconnected systems. In a school district facing high rates of chronic absenteeism among Pre-K and Kindergarten students, our team analyzed data from more than 3,600 students across 30 schools using multi-level logistic regression to examine how student characteristics and school contexts jointly shape attendance patterns. This approach revealed patterns that would be missed in a single-level analysis: for example, while grade retention was associated with higher absence rates, that relationship depended on the broader school environment, and students’ residential context independently predicted absenteeism, pointing to geographic inequities that operate outside school walls.
Building on this framework, CERE’s Children, Families, and Communities Portfolio extended multi-level methods to examine kindergarten readiness and classroom composition, using longitudinal administrative data to show that roughly 10% of the variation in student test scores occurs between classrooms and that the share of Pre-K–experienced peers in a classroom produces measurable spillover effects on all students. In practical terms, increasing Pre-K participation not only improves outcomes for individual children but also raises achievement for their classmates, suggesting that traditional estimates focused only on individual gains substantially understate program impact.
Taken together, this work demonstrates CERE’s ability to integrate advanced statistical methods with policy-relevant questions, producing findings that help clients identify where interventions are most effective, allocate resources with greater precision, and make stronger, evidence-based cases for program directions.