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

[Kit Baum <baum@bc.edu>]

(1) Try the estimation using xtivreg2. If you have anywhere near 2000  
variables in the dataset, -preserve- and drop some of those that are  
not in the equation to be estimated.

(2) "Hausman test easily rejects the fixed effects GLS."  A Hausman  
test rejects when the "efficient" estimator is found to be  
inconsistent. In the RE/FE setup, RE is "efficient" under the null  
that X \perp u and inconsistent otherwise. If the Hausman test is  
rejecting anything, it is rejecting that the random effects estimator  
is consistent.

It is not surprising that you could reject reject pooled OLS, which  
ignores unit-level heterogeneity, in favor of a FE model which  
incorporates it. But whether you can assert that X \perp u in that  
model is another issue.

Kit

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 21, 2008, at 02:33 , statalist-digest wrote:


> Thank you. As I mentioned I used -xtreg- for the
> original equation as I believed that random-effects
> GLS is a less biased estimator than OLS, since the
> data are grouped across workplaces (Moulton, 1987). A
> Hausman test easily rejects the fixed effects GLS.
>
> I tried to perform the Hausman test by estimating the
> model with -xtivreg- using re. But this is what I got:
>
>  . xtivreg pay  yrsed  (b4a2 = dhat) mnedwk2 e1a1
> p1a2-p1a5 v2a1-v2a5 v14a1 dis2 h2a2 h2a3 h4a2 h3a1
> h3a3 h3a4 f1a1 z2-z6  ratiopk2 n1-n4 n6-n12
> fnonmana1-fn6  fno1-fno6 ratiofk2 aga1  k2-k4 in1 c1,
> re
>
> no room to add more variables
*
*   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/