Instrumented JWDID

[Nay Iturbe]

Does anyone knows if JWDID can be implemented also by adding IV? And if the answer is yes, how can I do it? I am working with a staggered treatment, but I suspect endogeneity. The treatment is not reversible, and neither the instrument I want to use.

[FernandoRios]

Hi Nay,
JWDID cannot handle that directly. However, you could do a two step procedure.
1) use, say regress to run the DID for the endogenous variable and the instrument.
2) get the residuals
3) add residuals as explanatory variable to the main model.

I think that would be the equivalent to CF approach with DID. However, it will depend on you to justify that

[George Ford]

You can get JWDID results using reghdfe (or xtreg) if you create all the necessary variables. I suppose you could build a model using ivreghdfe, but I have no idea if that would be theoretically or computationally legit.