0
$\begingroup$

I wish to perform an IV regression where the first stage is estimated without fixed effects, but the second stage is estimated with fixed effects. Is this possible? Are there any obvious issues with this approach? Are there papers in which this has been performed before?

Improve this question
$\endgroup$
3
  • $\begingroup$ Could you provide more details? Are you reluctant to use fixed-effects in the first stage because the IV is time-invariant? $\endgroup$
    – Durden
    Commented Nov 4 at 15:15
  • $\begingroup$ My second stage has both country (this is my cross-sectional variable) fixed effects and time fixed effects. However, I think the estimation of my first stage does not require time fixed effects since it should not be influenced by time-related factors. $\endgroup$
    – Anonymous
    Commented Nov 4 at 16:26
  • $\begingroup$ IV logic requires that all exogenous regressors should also be included in the first stage. So, even if you feel you do not need the effects in the first stage, why not still include them? $\endgroup$
    – Christoph Hanck
    Commented Nov 5 at 12:28

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.