Andrea,
Thanks a lot for your reply.
Yes, noconst made the xtabond results much closer to the xtabond2
results. I realized after a while that, by default, xtabond includes a
constant in the fd equation, while xtabond2 doesn't.
However, with my dataset (an unbalanced panel), there is another source
of discrepancy: xtabond and xtabond2 don't handle collinearity in the
instruments the same way. The respective developers are looking at this
issue.
Jean Salvati
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of sistoand80
> Sent: Wednesday, August 18, 2004 9:41 AM
> To: statalist
> Subject: Re:st: Xtabond2 versus xtabond
>
> Dear Yontcheva Boriana,
> I think results differ because xtabond2 drop constant term. Try
>
> xtabond y x1...x4 time dummies, lag(1) noconst
>
> xtabond2 y L.y x1..x4 time dummies, noleveleq ivstyle(x1-x4
> time dummies, passthru) gmmstyle(L.y)
>
> I've successfully replicated xtabond results trough xtabond2
> but, as I don't have my results saved at University, I'm not
> sure at 100% about this procedure. If this specification is
> not the correct one, tomorrow I'll give you a more
> appropriate suggestion.
> Bye
>
> Andrea Sisto
>
> ---------- Initial Header -----------
>
> From : [email protected]
> To : [email protected]
> Cc :
> Date : Fri, 13 Aug 2004 14:31:21 -0400
> Subject : st: Xtabond2 versus xtabond
>
> > Hi all,
> > I'm new to stata and statlist. I have been trying to
> replicate results
> > obtained through using xtabond command with the xtabond2
> command, by
> > using the "noleveleq" option. I thought this option makes xtabond2
> > give a first difference estimator equivalent to the Arellano Bond
> > estimator obtained by using the xtabond command.
> >
> > My model is as follows: y= y(t-1) + x(t) + time dummies +
> individual
> > effects+ time varying error term.
> >
> > Where x(t) are strictly exogenous variables.
> >
> > In order to estimate this through Arellano Bond I use the following
> > command:
> >
> > xtabond y x1...x4 time dummies, lag(1)
> >
> > In order to estimate it through xtabond2 I use the
> following command
> > with the noleveleq option:
> >
> > xtabond2 y L.y x1..x4 time dummies, noleveleq ivstyle(x1-x4 time
> > dummies, passthru) gmmstyle(L.ed)
> >
> > Can someone please let me know why these two don't give me the same
> > results?
> >
> > Thanks a lot
> >
> > *
> > * 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/
> >
>
>
> *
> * 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/
>
>
*
* 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/
