st: reg3 in a panel context

[Kit Baum <baum@bc.edu>]

Mehmet,

I presume in mentioning reg3 you refer to estimation of simultaneous  
equations. The 'third stage', or Zellner step, of a 3SLS estimator  
only gains efficiency over single equation (limited information)  
estimation, so is never required to generate consistent estimates.  
Nevertheless, you may want to use a systems estimator in order to  
impose cross-equation constraints on the structural equations or test  
hypotheses.

That said, just as you can run 'regress'  or 'ivreg' / 'ivrge2' on  
panel data without taking the panel structure into account, you can  
run 'reg3' to perform pooled cross section/time series estimation of  
a set of structural equations. If one of the dimensions of the panel  
is short, and you want to perform the equivalent of a fixed effects  
estimator (assuming that you do not have dynamic equations), you  
could insert the appropriate set of dummies.

Alternatively, if you have a panel of e.g. a few units (say, the G7  
countries, or several two-digit SIC industries) over a long time  
span, you could define each country's data as a separate structural  
equation, and apply cross-equation constraints where theoretically  
appropriate. Imagine that you are estimating the demand and supply of  
internationally traded widgets as a two-equation system (with Q and P  
as the endogenous variables), and you have data for the G7. You could  
then set it up as a system of seven demand equations and seven supply  
equations, and run reg3 on that 14-equation system. To carry this out  
the data would have to be in the 'wide' format.


Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


On Aug 9, 2006, at 3:58 AM, Mehmet Eris wrote:


> Dear Dr. Baum,
>
> I have a question about a Stata module. I currently need a stata
> module which can do what reg3 does using panel data. Does such a
> module exist? I searched for it for a while and it seems that some
> other people were looking for it as well. I am wondering if you know
> how I can go about this. Thank you very much for your time!
>
> Best regards,
> Mehmet
*
*   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/