<>
I am estimating a model with multiple endogenous variables (i.e., 3)
using xtivreg2. The output includes the Hansen J. Why is that we only
have 1 Hansen J (instead of 3) when there are 3 endogenous variables?
My intuition tells me that that is because we are simply testing
whether the instrument set is correctly excluded from the second stage
model (which is only 1). Can anyone suggest any references where I
could get more intuition on the issue?
As discussed in Baum-Schaffer-Stillman papers in Stata Journal 2003,
2007, IMEUS and the Stata manual, the Hansen-Sargan statistic is a
measure that pertains to the overall regression. In the IV-GMM
context, it is a test of the 'overidentifying restrictions' that each
of the instruments are orthogonal to the error process. Thus it does
not matter how many endogenous variables appear in the equation, and
in fact we are not questioning their asserted endogeneity. What
matters, for the Hansen J dof, is the number of OID restrictions. You
can challenge the endogeneity/exogeneity status of endogenous
variables in B-S-S -ivreg2- (available from ssc) using the endog() or
orthog() options.
Kit
Kit Baum | Boston College Economics & DIW Berlin | http://ideas.repec.org/e/pba1.html
An Introduction to Stata Programming
| http://www.stata-press.com/books/isp.html
An Introduction to Modern Econometrics Using Stata | http://www.stata-press.com/books/imeus.html
*
* 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/
