Heckman 2 steps model variables selection

[Quentin Richard]

Hello,
I'm trying to develop a heckman 2 steps model in Stata and I have a very simple question.
Can the variables used to predict the fact to be selected (after the "select" part of the command) be totally different than the dependent variables included in my main equation of interest? Is it correct or should the variables of the main equation also appear in the probit of selection?
Thanks a lot for your help,
Best,
Quentin

[Maxence Morlet]

"Can the variables used to predict the fact to be selected (after the "select" part of the command) be totally different than the dependent variables included in my main equation of interest?"

Short answer is yes

You may also want to select an instrument, which will appear in the selection equation but not in the outcome equation.

Please read this article: https://deepblue.lib.umich.edu/bitst...pdf?sequence=1

on the importance of instrument selection in Heckman models.

[Quentin Richard]

Thanks a lot Maxence!
Best,
Quentin