(Visualizing) Plausible Treatment Effect Paths

Freyaldenhoven, Simon; Hansen, Christian

doi:10.1257/aeri.20250442

Abstract

We study inference on dynamic treatment effects. We propose simple additions to standard visualizations aimed at enhancing their informativeness. First, we recommend reporting joint hypothesis tests and summary measures that are often of direct interest alongside the standard event-study plot. Second, we introduce a data-driven procedure that estimates a smooth approximation to the treatment-effect path within a prespecified, economically motivated class and produces corresponding confidence regions. The resulting intervals can be substantially tighter than conventional uniform bands while maintaining formal coverage guarantees through post-selection inference (Berk et al. 2013). The accompanying R package plausibounds implements our suggestions.

蝌蚪网

蝌蚪网 Review: Insights