Re: st: 3 simultaneous equations with large amount of dummies

[Kit Baum <baum@bc.edu>]

<>
Along the lines of many postings, unless you have cross-equation  
restrictions, there is nothing wrong with estimating each equation of  
this system separately using a limited-information estimator.  It is a  
strangely common misconception that systems of linear equations must  
be estimated with -reg3-. -xtivreg2, fe- (findit xtivreg2) will handle  
the fixed effects and give you consistent estimates of each equation's  
parameters if the equations and instrument sets are properly specified.

The only thing a systems estimator can give you is efficiency (and a  
system with > 9000 parameters will not be a pretty sight). The  
downside of systems estimation is that one misspecification error will  
pollute estimates of all equations. For these reasons (as well as  
convenience) a single-equation estimation approach has much to  
recommend it.

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



On Mar 25, 2009, at 02:33 , xinzheng wrote:

> (1) I am estimating a simultaneous equation system with 3 equations,  
> but I need to include firm fixed effects in each equation. If I just  
> use reg3 and add firm dummies, there are two many firms (>3000), it  
> takes a very long time and sometimes I could not get the results. I  
> was wandering what kind of other STATA command I can use.
>
> (2) I am running a IV estimation with a firm level fixed effect, if  
> I add firm dummies, there are more than 3000, which makes it  
> impossible to get the results. Is there any other STATA command I  
> can use?

*
*   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/