st: RE: Random Effects model with Dynamics

[Rodolphe Desbordes <rodolphe.desbordes@strath.ac.uk>]

Dear Albert,

A dynamic random effects estimator is inconsistent. On the other hand, a "System-GMM" estimator would be a consistent estimator.

Rodolphe

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Lee, Albert
Sent: jeudi 25 février 2010 15:42
To: statalist@hsphsun2.harvard.edu
Subject: st: Random Effects model with Dynamics

My dataset is an annual panel with many subjects (tens of thousands) and relatively short time period (at most 10 years).  This panel is very unbalanced (T ranges from 1 to 10).  Looking at some summary statistics, I'm convinced that it is best to model this as a random effects model with a lagged dependent variable as a regressor, i.e.,

y(i,t)=a+b*y(i, t-1)+e(i)+u(i,t),

where e(i) is the subject random effects and u(i,t) is the noise.

I'm aware of the xtmixed and xtabond commands in stata.  But the model as specified may need a combination of the two estimation methods.

Questions:
1.  Does stata provide a routine that estimates dynamic model with random effects?
2.  Theoretically, is this an overkill to model dynamics and random effects in the same model?

Extra credit:
1.  I'm in fact estimating several equations like the one indicated above.  Is there a seemingly unrelated technique that I could use to exploit the correlations of the error terms from these equations.  I guess that it will be a simultaneous random effects panels with dynamics.  Any thoughts?

Many thanks in advance,

Albert Lee

*
*   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/

*
*   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/