On Feb 23, 2005, at 2:33 AM, Mark wrote:
I don't recall seeing anything, though,
about IV being less prone to weak instrument problems than (2-step)
GMM.
But this literature is expanding rapidly, and maybe there's something
out
there that discusses this.
Since weak instrument problems are related to poor correlations between
instruments and the regressors, it would not seem that any approach
will be able to overcome those weak associations between the variables.
It is certainly true that standard IV (2SLS) and IV-GMM use the
instruments in different ways: the 2SLS approach creates linear
combinations of the instruments while the GMM approach retains their
individual identities. But since correlation is a linear operator, if
the equation is 'essentially unidentified' with the individual
instruments, it will have the same 'essential unidentification' if
those instruments are combined linearly.
Kit
*
* 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/
