Student Growth Percentiles: Testing for Validity and Reliability

• In Event: 33.043 - The Science of Teacher Evaluation: Implications for Practice

Abstract

An ongoing debate among researchers, policy-makers, and practitioners about the rationale for using statistical models for teacher evaluation has fueled extensive examination of value-added models (VAMs). However, the student growth percentiles model (SGP), which is being used more often than any VAM, has received significantly less attention. In this study, researchers fill this gap and evaluate validity and reliability of SGP estimates using the data from Arizona school districts collected over three years. Researchers found that the SGP model has low power to identify best and worst performing teachers. Moreover, teacher growth estimates and observational score are only weakly correlated which raises concerns about the overall validity of the SGP model as the only means to evaluate teacher performance.

Authors

• Margarita Pivovarova, Arizona State University - Tempe

• Audrey Amrein-Beardsley, Arizona State University