st: Standard errors in first stage ivregress vs. reg

[Ingo Isphording <Ingo.Isphording@rub.de>]

Dear statalist,

I want to reproduce the first stage of ivregress by using reg to get
access to the coefficient vector and covariance matrix. Unfortunately,
the standard errors differ between

ivregress 2tsls y x1 (x2 = z), robust firststage

and

reg x2 x1 z, robust

The point estimates are the same, the standard errors differ
(marginally). This problem only occurs if I estimate the model with
robust standard errors.

To make it clear, I only refer to differences in the first stage,
which I do not understand. Is there any other way to obtain the
estimates to use them in estout?
*
*   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/