st: Re: fixed effects vs random effects

[Kit Baum <baum@bc.edu>]

Not so. xtreg,fe is the least squares dummy variable model: OLS with  
unit-specific dummies. If OLS is inappropriate for that model due to  
endogeneity of one or more regressors, then taking care of the  
unobserved heterogeneity does not make it consistent. Use xtivreg or  
better yet xtivreg2 using an appropriate set of instruments. xtivreg2  
has extensive diagnostics for the validity and strength of  
instruments. It does not, however, deal with random effects models.  
But if your xtreg,fe has endogeneity, there's no way that xtreg,re on  
the same model would be appropriate either, as it assumes that the  
unit-specific random terms are also independent of the idiosyncratic  
error.


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


On Feb 2, 2007, at 2:33 AM, Shams wrote:


> In this regard, my understanding is that Fixed Effect
> (FE) estimation (with xtreg, fe) would control for
> both unobserved and endogeneity in the regressors and
> consistent and efficient for large T and if there is
> sufficient within panel variation. Alternatively,
> random effect (RE) estimation (xtreg, re) would only
> control for random unobserved heterogeneity but not
> the endogeneity issue.
*
*   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/