Lacking Predictiveness of Success and Propagated Biases: The Use of Summative Assessments in Mathematics Course Placement Decisions

• In Event: Systemic Factors That Limit English Learners' Access to Mathematics: Lessons From a Research-Practice Partnership

Abstract

Results from large-scale standardized educational assessments are often used for purposes other than those that informed their design. One such example is the use of large-scale summative assessments when making mathematics course placement decisions for accelerated pathways in middle school and during the transition from middle to high school—most critically for Algebra I. Additionally, there is evidence of systematically lower performance by English Learners (ELs) in many such assessment systems. Using administrative and transcript data from a multi-year research-practice partnership focusing on helping school districts better support their ELs, I advocate against the extant, yet inappropriate use of California Assessment of Student Progress and Performance (CAASPP) scores to determine mathematics course access. Specifically, (i) I describe the relationship between middle school CAASPP scores and subsequent mathematics high school success through a series of overlapping histograms; (ii) I report results of logistic regression analyses, showing how poorly CAASPP scores predict passing select high school mathematics courses; and (iii) I offer explanations of some of the processes at play. Results suggest that using students’ large-scale summative assessment performance to determine (partially or fully) access to mathematics courses is problematic for two reasons: (i) mastery of prior year’s learning are conceptually different from aptitude for specific future learning or preparedness for a certain mathematics course; (ii) potential (linguistic) biases in these assessments may propagate and result in discriminatory mathematics course placements. Findings are of relevance to both school districts still using large-scale assessment performance as a factor in course placement decisions, as well as to researchers working with educational assessment data.

Author

• Julian Maximilian Siebert, Stanford University
  Presenting Author