Hi all,

I am working on panel data and I assume that at least one of my explanatory variables is endogenous, but I think that I can account for the endogeneity by either using fixed effects or IVs. My approach seems a bit ad hoc, but the idea is to include a lagged dependent variable (LDV) as regressor and create a Nickell bias. When I get large and significant estimates for the LDV in my simple model, I would then add different fixed effects to my model and maybe the estimates for the LDV get smaller. If this is the case, the chosen approach could therefore account for the endogenous nature of my variable to the extent of the reduction of the effect of the LDV. So the idea is to use the Nickel bias as a test for the usefulness of my fixed effects.

Do you have any suggestions on this?

Best,
Marius