Basically, would you like a graph like that on the left (theory driven sample restrictions and controls) or on the right (data driven weighted controls seeking to "impose" pre-trends)?
(Note - in paper, I will isolate main estimate in its own figure. These would be for robustness checks)
Atheendar Venkataramani
Help me rewrite my paper!
Would you rather see:
1. Noisier (with mostly flat pretrends) in a standard TWFE with sample restrictions that follow history/theory
2. Flat pretrends + more precisely estimated coefs in Synthetic DID
(Refs seem to be skeptical about 2. Baseline is hard bec noise for 1)
