Dear Sirs:
Can anyone tell me if -xtabond2- can be used to estimate a model where one
of the regressors is endogenous and one is predetermined (the lag of another
endogenous variable, not the outcome). For example, I am interested in
estimating something akin to the following system of equations:
(1) P_ij = U_ij + C_i(j-1) + X ,
(2) U_ij = P_ij + C_i(j-1) + Z, and
(3) C_ij = U_ij + P_ij + Y
where i indexes subject; j indexes time; P, U, and C are endogenous
variables; and X, Y, and Z are exogenous variables. I think (though I am
not certain) that I can use -xtabond2- to estimate each equation separately
to overcome the consistency issue associated with lagged endogenous
variables.
For example, I think that the following code would be appropriate for
estimating equation (1):
xi: xtabond2 P U L.C X , gmm(U, lag(2 . )) gmm(L.C) iv(X Y Z,
eq(level)) twostep h(1) robust
Assuming -xtabond2- is capable of doing what I want, could someone please
tell me whether or not the above code is correct?
Thank you very much.
Best Regards,
--
James Shaw
College of Pharmacy
The University of Arizona
*
* 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/
