Dear all,
I'm writing from the department for public policy of the University of Eastern Piedmont (Alessandria, Italy). I've experienced some problem when I've tried to replicate xtabond results on xtabond2. In my dynamic regression i used cross dummy variables. As first difference lead to drop cross dummy variables due to collinearity problem, i specified this variables in diffvars list. As from xtabond help file, diffvars(varlist) specifies a set of variables that have already been differenced to be included as strictly exogenous covariates. When I tried to replicate this with xtabond2 i experienced some problems as dummy variables has been dropped in first-differenced equation.
I've the following dynamic panel
Yit = Yit-1 + Xit + vi + eit + cross_dummy,
where Xit is a vector of strictly exogenous variables.
I performed this xtabond command: "xtabond Yit Xit, diffvars(cross_dummy) noconst"
Then I've typed this xtabond2 command (only FD equation): "xtabond2 Yit l.Yit Xit, iv(Xit, eq(both)) gmm(l.Yit cross_dummy) noleveleq"
While I replicate when I've dropped cross dummies, when I've specified the model including dummy variables cross dummies have been dropped due to collinearity. How can I specify that this variable s have to be taken as they are, without first-differencing? A further problem is that when i drop "noleveleq" command, trying to perform a system of GMM, cross dummy variables appear in final regression but I think results could be biased as cross dummies have been dropped in Fd eq.
Could anyone help me?
Best regards
*
* 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/
