st: Re: 2SLS with probit in the first stage regression

[Kit Baum <baum@bc.edu>]

I guess I don't understand your setup. You're using xtreg, which is  
appropriate for panel data. If the original regression is

.xtreg y training meanworkplacetraining x2 x3 x4

and there is a possibility that training is endogenous, you  
definitely are violating the key assumption of the random-effects  
model, that the regressors are distributed independently of the  
errors. You should be estimating this equation allowing for  
endogeneity of the suspected variable, and need to perform a Hausman  
test of the appropriateness of the key assumption. You can do the  
Hausman test by estimating the model with -xtivreg- using RE and FE.  
If you find that the RE assumption cannot be maintained, then use - 
xtivreg2- with fixed effects and use the -endog()- option to test  
whether the endogeneity is an issue.


Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


On Feb 16, 2008, at 2:33 AM, statalist-digest wrote:


> Thank you, Kit. I should have mentioned also that my
> original equation which has one of the RHS regressors
> as potentially endogenous is a random effects GLS
> model. Also I should have pointed out I am using
> cross-section data. Would your kind answer still
> apply?
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/