>>I have a panel dataset for a sample of 20 countries over 8 periods
>>I`ve use GMM method to estimate a model of convergence like this
>>yit=b yit-1 + q xit + uit where xit are exogenous variables I could be
>>posible that difference GMM work better than sysyem GMM?
To go back to the original question, you need to define what you mean by "better". Are you concerned about efficiency, consistency, small-sample bias, ...?
In this respect, the only thing that the theory tells you is that, if all your orthogonality conditions are valid, then both estimators are consistent, and system-GMM is more efficient than first-difference GMM.
Now, if the orthogonality conditions for the first-difference equation are valid, but those for the level equation are not, then system GMM is not better than first-difference GMM in any way.
This can happen, for example, if the regressors used in the orthogonality conditions for the level equation are correlated with the individual effect. If the p-value for the J-stat is smaller with system GMM than with difference GMM, then that's probably what the problem is.
As far as I know, not much is known about the small sample-bias of the two estimators, and all the information that we have on this topic comes from Monte-Carlo simulations. The theory doesn't tell you anything in that area. Other listers may have more information about that.
Jean Salvati
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> [email protected]
> Sent: Thursday, February 09, 2006 7:01 AM
> To: [email protected]
> Subject: RE: st: GMM diff vs sys
>
> Giovanni and Nick
> Thanks a lot
>
>
> > The finite-sample bias approximations used by xtlsdvc requires an
> > estimate of the regression variance. So the estimator is
> implemented
> > into two steps:
> > 1) get the variance estimate from residuals obtained from some
> > *initial* estimator
> > 2) work out bias approximations and correct LSDV
> >
> > An help file is available with xtlsdvc (-help xtlsdvc-),
> but you might
> > also want to have a look at my Stata Journal paper (and the
> references
> > therein):
> >
> > "Estimation and inference in dynamic unbalanced panel-data
> models with
> > a small number of individuals,"
> > Stata Journal, StataCorp LP, vol. 5(4), pages 473-50
> >
> >
> > Giovanni
> >
> > [email protected]:
> >
> >> The error is this
> >>
> >> xtlsdvc lnrgdpl lnrgdpli lnts lnngd
> >> option initial() required
> >> r(198);
> >>
> >> but i don`t know what is an initial option I need a
> regression of a
> >> convergence model like this lnYit= a + b lnYit-1 + d Xit + ni + vt
> >> +uit where Yit is per capita gdp and Xit are population growth and
> >> interest rate
> >>
> >> Thanks
> >>
> >>
> >> where: - level of per capita income for i'th country in t'th
> > period.,
> >> Xi,t - matrix of the variables, which represent the
> steady state for
> >> i'th economy (in logarithms),
> >> - individual effect of i'th country,
> >> - time specific effect of t'th year,
> >> - independent stochastic error term.
> >>
> >>
> >> where: - level of per capita income for i'th country in t'th
> > period.,
> >> Xi,t - matrix of the variables, which represent the
> steady state for
> >> i'th economy (in logarithms),
> >> - individual effect of i'th country,
> >> - time specific effect of t'th year,
> >> - independent stochastic error term.
> >>
> >>
> >> > -xtlsdvc- requires an -initial()- option.
> >> >
> >> > Otherwise, to quote the FAQ:
> >> >
> >> > "Say exactly what you typed and exactly what Stata typed
> (or did)
> >> > in response. N.B. exactly! If you can, reproduce the
> error with one
> >> > of Stata's provided datasets or a simple concocted
> dataset that you
> >> > include in your posting."
> >> >
> >> > Nick
> >> > [email protected]
> >> >
> >> > [email protected]
> >> >
> >> >> I`ve tried this
> >> >>
> >> >> xtlsdvc depvar varlist
> >> >>
> >> >> But It hasn`t funtioned Could you tell me what it`s wrong?
> >> >>
> >> >
> >> > *
> >> > * 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/
> >>
> >
> >
> > --
> > Giovanni S.F. Bruno
> > http://ideas.repec.org/e/pbr136.html
> > Istituto di Economia Politica, Universit� Bocconi Via U. Gobbi, 5,
> > 20136 Milano Italy tel. + 02 5836 5411 fax. + 02 5836 5438
> > *
> > * 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/