IV Fixed Effects Model combined with Heckit

[Gonzalo AresdePargaRegalado]

Dear Statalist,

Following this discussion:

https://www.statalist.org/forums/for...quations-model

I want to know if it is possible to estimate an IV Fixed Effects Model combined with a Heckit. Is there any command to do this?

Thanks a lot!

[FernandoRios]

Yes, there are two options.
1. Stata 16 has a new set of commands " xteregress " that estimate random effect models, and can account for fixed effects and selection.
2. Search for panel effect and heckman in the forum. Prof Wooldridge has worked on a paper and his coauthor has the code that also deals with fixed effects and selection.
HTH
Fernando

[Gonzalo AresdePargaRegalado]

Thanks a lot for your help, Fernando.

Maybe I am mistaken but "xteregress" does not have a FE option, isn't it?

I found the resources that you mentioned in your reply. I read rapidly the codes but I found out that process 4.1 (detailed here http://myweb.fsu.edu/asemykina/selectionPS.pdf) is not available. Do you know of any application (with code available) using a IV Fixed Effects considering the selection with a Tobit or a Heckit?

I have an additional doubt:

Should the panel be balanced for the indicator?

[FernandoRios]

You are correct, xteregress option only estimates Random effects models.
for the FE estimator, here is the code they have posted http://myweb.fsu.edu/asemykina/two_s..._parametric.do
In terms of the application itself, it will require some careful work. As this is a step by step application. There is no "ado" file that does this.
THe bootstrap version is relatively straight forward, but again, you may need to read their paper carefully before adapting and implementing the estimator
For your last question, I'm unaware of that requirement. Perhaps Jeff Wooldridge may join this and give you a better suggestion.
Fernando

[Gonzalo AresdePargaRegalado]

FernandoRios I read Estimating panel data models in the presence of endogeneity and selection: https://www.sciencedirect.com/scienc...25?via%3Dihub#!

The authors have procedure 4.1 which is something like what I was looking for. Nevertheless, I wonder if Jeff Wooldridge or River Huang have any idea if an ordered probit may be used instead of a probit. If so, the procedure and assumptions would be the same?