I have a model $Z = \alpha X + \beta Y + \gamma C + \epsilon$. I am interested in the relative effects of $X$ and $Y$ on $Z$. However, $X$ and $Y$ are endogenous. I have identified two instruments $X'$ and $Y'$ respectively. $C$ is an exogenous control variable (fixed effect).

How do I go about estimating the model with these two instruments for the two endogenous variables? Do I need to use 3SLS? I am fairly new to this field. Any resources that address this would be very helpful, thank you.

  • 1
    $\begingroup$ No need for 3sls, you can just run 2sls as you normally would. $\endgroup$ – BB King Aug 16 '18 at 14:49

I don't think you need anything like a 3SLS. Just run the first stage for the two endogenous variables separately:

$$\tilde{X}=\alpha_1+\beta_1X'+\beta_2Y'+\beta_3C+\epsilon$$ $$\tilde{Y}=\kappa_1+\gamma_1X'+\gamma_2Y'+\gamma_3C+\delta$$

Then run a regression of Z on the predicted value of $\tilde{X}$, $\tilde{Y}$ as and $C$. If the assumptions are all satisfied, this should give a consistent estimate of the coefficients.

I also found this thread online. It might be helpful. https://www.stata.com/statalist/archive/2012-08/msg01238.html

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.