st: what does Stata do with the fixed-effect in Xtlogit

[Milet Emmanuel <emmanuel.milet@cepii.fr>]

Dear statalist users,

I have a simple interrogation regarding the Xtlogit procedure:
what does stata actually do with the fixed effect ? In an xtreg it 
subtracts the mean, but I could not find in the xtlogit.ado any clue 
about what's going on here.

I wonder since when I run the two following estimations, my results are 
somehow different:

Estimation 1: xtlogit y x1 x2, fe i(panel_var)
Estimation 2: xi: logit x x1 x2 i.panel_var

I would rather trust the Estimation 2, but I wonder why the results are 
different.

Here is an abstract of the results, for you to see the differences:

With the Xtlogit:
------------------------------------------------------------------------------
     xstatus |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
       lgdpc |   1.135492   .1741228     6.52   0.000      .794218    
1.476767
        lpop  |   .9073863   .0665109    13.64   0.000     .7770274    
1.037745
       ldist    |  -.9110466   .0912266    -9.99   0.000    -1.089848   
-.7322458

With the Logit including fixed-effects:
------------------------------------------------------------------------------
     xstatus |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
       lgdpc |   1.176038    .178538     6.59   0.000     .8261099    
1.525966
        lpop  |   .9565665   .0688508    13.89   0.000     .8216214    
1.091512
       ldist    |  -.9724132    .093985   -10.35   0.000     -1.15662   
-.7882059


Thanks,

Emmanuel Milet
CEPII, Paris, France
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/