0
$\begingroup$

I'm working on a a bivariate logistic regression. But I have an endogeneity problem and I want solve it through 2sls with 2 instrumental variables.

My thought was to regress (OLS) in first stage and later a bilogit in second stage. Is this a suitable analysis? Is it a "forbidden regression"? What would be the best option that helps to solve it?

Improve this question
$\endgroup$
6
  • 2
    $\begingroup$ If your only question is about (the existence of R) code, it is off topic here. If you have a statistical question about 2sls when 1 stage is a binary variable, please edit to clarify. $\endgroup$ – gung - Reinstate Monica Aug 11 '17 at 15:36
  • $\begingroup$ I see what you mean. Ok. On other hand, to carry out a 2sls with ols in 1st stage and logit in 2nd is a "forbbiden regression", isn't it? thanks for you answer, gung. $\endgroup$ – SMD Aug 11 '17 at 15:49
  • $\begingroup$ If you have a statistical question about 2sls when 1 stage is a binary variable, please edit your question to focus on that. Otherwise, this will end up being closed. $\endgroup$ – gung - Reinstate Monica Aug 11 '17 at 16:00
  • $\begingroup$ I've edited question, gung $\endgroup$ – SMD Aug 11 '17 at 16:17
  • $\begingroup$ This question is now on topic here. I'm voting to leave open. $\endgroup$ – gung - Reinstate Monica Aug 11 '17 at 17:14
1
$\begingroup$

Short answer : No, you can't. That is indeed forbidden regression what you're trying to do.

Possible solutions : some brief discussion, google the keywords to learn more.

  1. Use LPM(linear probability model) - just use regular 2SLS with your data. sounds crazy, but a lot of people do this in economics literature. Make sure that you get the standard errors right.
  2. MLE - specify the distribution of your first stage equation, jointly with the second stage binary outcome equation. (i.e. use probit model with bi-variate normal errors)
  3. Control function approach - basically add the residuals from the first stage equation to covariates in the second stage.

for details, see Blundell & Powell(2004) or Rivers & Vuong(1988).

Improve this answer
$\endgroup$
1
  • $\begingroup$ If I could, I would rate your answer as useful. Thank you so much $\endgroup$ – SMD Nov 17 '17 at 20:46

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.