Addressing Measurement Error Bias and Omitted Variable Bias in Modeling Compositional Effects in Educational Settings

• In Event: Novel Approaches for Model Fit

Abstract

The main objective of our study is to quantify the impact of failing to account for measurement error bias and omitted variable bias in the school/class compositional effect estimate. We revisit a database used in the past for the assessment of the peer effect in education, and we find that the small, positive and statistically significant effect associated with school average achievement becomes less positive and non-significant once adjustments for measurement error are made. Our findings question empirical results of past and current research evaluating the peer effect based on sub-optimal models, that fail to control for measurement error. The proposed methodology can be relevant to debates on test-based value-added measures of teacher performance, parental choice and selective schooling.

Authors

• Ioulia Televantou, University of Nicosia Research Foundation & Theological School of Cyprus

• Herbert W. Marsh, Australian Catholic University