A Monte Carlo Simulation for Examining the Performance of Negative Outcome Control Estimator as a Generalization of Difference-in-Differences under the Violation of Identification Assumptions

• In Event: Division D Graduate Student in Progress Research Gala

Abstract

ANCOVA and difference-in-difference are standard methods for analyzing data from the pre-post design. A recent study showed that the pre-post design is a special case of negative outcome control (NOC) designs and that we can estimate the treatment effect with the NOC estimator under less restrictive assumptions. Although the NOC may be attractive to educational researchers, the properties of the estimator under the violation of assumptions are not examined sufficiently. Therefore, this study investigates the consequences of the violation through a simulation. The results showed that NOC estimates could be biased, depending on the existence of outcome-specific confounders.

Author

• Kazuki Hori, Benesse Educational Research and Development Institute