Effect of Summer Learning Loss on Aggregate Estimates of Student Growth

Abstract

Recent reforms in federal educational policy now mandate the use of student assessment data to evaluate teachers and principals. Despite the widespread adoption of Student Growth Percentiles (SGPs) and other models to link student achievement growth to teacher and school effectiveness, little research exists evaluating the validity of the resulting effectiveness estimates for use in high-stakes personnel evaluation systems. This paucity in the literature is especially problematic given that significant correlations between effectiveness estimates and student characteristics, specifically poverty, have been well documented. This dissertation explores summer learning loss as one potential source of bias in annual estimates of student growth for teacher and school evaluation. The guiding hypothesis is that economically moderated summer learning patterns are contributing to systematic error variance in teacher and school effectiveness estimates when calculated based on annual test scores. Datasets from two, nationally distributed commercial interim assessment programs are analyzed separately and their results discussed. Results reveal that the extent of summer learning loss, and by extension, its effect on the validity SGPs for evaluation purposes varies by subject, grade level, and testing program. Statistically significant correlations between mean Student Growth Percentiles and summer learning loss range from r = -.310 to r = -.662. Implications for fairness and education policy are discussed.