Long Differences, Short Gaps, and the Principle of Falsification in Event Studies

This is part 3 in a series I started back in December 2024 on event study construction under differential timing using tools like csdid, did_multiplegt, and did_imputation. You can read part 1 here and part 2 here. In this post, I want to make a straightforward argument: when you’re plotting event studies, you should use long differences, not short gap…