I've seen questions about few treated units/few clusters appear frequently here. Lee & Wooldridge (2025) offer an estimation approach in such conditions, though modern methods like sdid are also useful.
I've posted on SSRN how to implement the LW approach for both common and staggered timing. It's quite straightforward, especially when using asreg, as I do.
I'm sure there's other approaches that might be tighter, but my proposal only requires a few lines of code.
Common Timing:
https://papers.ssrn.com/sol3/papers....act_id=5544598
Staggered Timing:
https://papers.ssrn.com/abstract=5643671
Any comments are appreciated.
I've posted on SSRN how to implement the LW approach for both common and staggered timing. It's quite straightforward, especially when using asreg, as I do.
I'm sure there's other approaches that might be tighter, but my proposal only requires a few lines of code.
Common Timing:
https://papers.ssrn.com/sol3/papers....act_id=5544598
Staggered Timing:
https://papers.ssrn.com/abstract=5643671
Any comments are appreciated.
