Placebo zones in discontinuity‐based designs: Estimation, inference, and implementation

Key Points

• The aim is to develop a new model-selection algorithm for regression discontinuity designs that accounts for placebo zones.
• Proposed a new model-selection algorithm assessing candidate models within a placebo zone.
• Evaluated the performance of models differing by bandwidth and other parameters.
• Outlined sufficient conditions for asymptotic optimality of the approach.
• Implemented the procedure through Stata commands and compared its performance with existing methods.
• The proposed method performed favorably in simulations calibrated to practical applications.
• Demonstrated asymptotic optimality under restrictive conditions and effective performance in broader scenarios.
• The randomization inference procedure showed improved accuracy with placebo estimates.

Abstract