Advancing Causal Inference in Education: Testing Instrument Exogeneity in Pretest–Posttest Designs

• In Event: AERA Roundtable Session Friday 3:45 pm - Gold 4
  In Roundtable Session: RT5: Advances and Implications for Causal Inference and Power in Randomized Treatment Designs

Abstract

Instrumental variable (IV) methods are essential for addressing unmeasured confounding in educational research, yet the validity of these methods depends critically on exogeneity assumptions that are rarely tested empirically. This paper introduces an innovative diagnostic approach to evaluate instrument validity in pretest–posttest designs. Building on structural parallels between IV estimation and compass variable–adjusted difference scores, we prove that discrepancies between these estimators reliably signal exogeneity violations. Monte Carlo simulations and reproduced analyses from empirical studies evaluating educational policies demonstrate the method’s practical utility. This approach strengthens causal inference rigor and supports more equitable, responsible educational policy evaluations.

Author

• Naram Gwak, Seoul National University