st: Standard errors with the vce(robust) option smaller than those without.

[Naomi Bosler <naomi.bosler@uni-konstanz.de>]

Dear all,

I am estimating a negative binomial model. As it is likely
that the observations are not independent from each other, I have
estimated my model using robust standard errors and something weird is
happening: The robust standard errors are SMALLER than the normal
standard errors. Some econometrics books say that this can occur,
although they admit that this is empirically rare. They do not say WHY
this might happen, which is what I need to find out.

Please note: I am not using the vce(cluster clustvar) option, but the  
vce(robust) option, therefore problems of negative correlation within  
clusters, as noted in FAQ:  
http://www.stata.com/support/faqs/stat/cluster.html are not applicable.

Thanks in advance for any help.
Best,
Naomi

*
*   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/