Disentangling Student Sorting Bias From Teacher Value-Added Estimates

• In Event: Issues in Value-Added Models for Teacher Accountability

Abstract

Recent studies found that teacher causal effects derived from Value-added models (VAMs) are generally biased due to variation in student composition, which is referred to as the student sorting bias. This is more severe when the teacher effects are estimated based on one year of data. This study investigates the extent of student sorting bias associated with the VAM scores for the Measures of Effective Teaching (MET) project and examines the effectiveness of experimental and non-experimental strategies for correcting such bias through analysis of the MET data of about 2,500 teachers from six large school districts during 2009-2011. Suggestions are provided on improvement of value-added models and related analytic approaches to deriving objective and reliable teacher effectiveness measures.

Author

• Ning Rui, Westat