I remember reading a paper where the Difference-in-Differences was a bit special, but I had forgotten the details. My recollection was
- Two time periods, specifically one time period before treatment and one time period after treatment.
- The "treatment" was such that all firms had to comply with a certain share of a variable X.
- The treatment group consisted of firms with very low (or zero) share of X before the treatment, and
- The control group consisted of firms with a share already close to the required share of X before the treatment.
The idea was that the treatment group received most of the shock, while the control group was little affected.
Do you know a paper that uses such a framework?