Examining Conditional Bias in Correlations Between Value-Added and Observational Scores

• In Event: The Fairness of Examining Teachers' Quality and Competencies

Abstract

We investigated whether nonlinear relationships between value-added and observational scores might be the result of bias, as related to student demographics. More specifically, we analyzed overall relationships and estimated correlations conditional on student demographics by looking at quartiles of teachers’ distributions who taught a) free-reduced lunch students, b) special education students, c) English language learners, and d) racial minority students. As per prior studies, we found that overall relationships between value-added and observational scores were weak-to-moderate, and we found the distinct correlations accounting for student demographics also indicated weak-to-moderate and often statistically insignificant results once disaggregated. These results imply that observed weak-to-moderate relationships between observational scores and value-added are unlikely to be driven by bias due to student demographics.

Authors

• Audrey Amrein-Beardsley, Arizona State University

• Margarita Pivovarova, Arizona State University

• Jeong Im Jin, Arizona State University