We consider estimation and inference for treatment effect paths. Examples include dynamic treatment effects, impulse response functions, and event study paths. We present two sets of plausible bounds to help visualize uncertainty associated with these paths. Both plausible bounds are often tighter than traditional confidence intervals and can provide insights even when traditional (uniform) confidence bands appear uninformative. Our first set of bounds covers the average (or overall) effect rather than the entire path. Our second set of bounds imposes data-driven smoothness restrictions on the treatment path, using post-selection inference (Berk et al. [2013]) to provide formal coverage guarantees.

Download the appendix.

View the Full Working Paper