Dear Stata users,
I want to apply a Heckman estimation to a dynamic panel data with sample selection using the newly available xtheckman command in Stata 16.
I have a CSTS panel with T=5 and N=100 and would like to check the robustness of my System-GMM estimation results using a lagged dependent variable in the model.

Code:
xtheckman y l.y  x  z  xz  , select(dummy= gdp demo)
Although the results are very similar to the GMM estimates I wonder whether it is valid to include a lagged dependent variable in the random effects linear regression model (with sample selection)?
Literature recommendations welcome!

Thank you very much in advance!
Best, Steve