Re: st: 3SLS with different instruments for different equations

[etanebay@yahoo.com]

Sorry if I'm being slow here, but I think the equations are exactly identified?:

Y1 is endogenous to Y2, so I used Z1 - this works in 2SLS

Y1 is also endogenous to Y3, so I used Z2 - this works in 2SLS

(Y2 and Y3 are not endogenous)

I'd like to model:

Y1 = Y2, Y3, Xi

But I would like to restrict Z1 to instrument Y2 only, and Z2 to instrument Y3. 

(Tried 3SLS but it does 1st stage Y2 with Z1 and Z2 and Y3 with Z1 and Z2 - and an earlier post advised gmm) 
 
Data are in panel and I have 3 moment equations:

Y1 = Y2, Y3, Xi
Y2 = Z1, Xi
Y3 = Z2, Xi

Can someone help with the syntax please? I tried the following but keep getting error messages

. gmm (Y1 - {b1}*Y2 - {b2}*Y3 - {b4}*X1 - {b5}*X2 - {b6}*X3 - {b0}) (Y2 - {b7}*Y1 - {b8}*X1 - {b9}*X2 - {b10}*X3 - {b11}*Z1 - {b12}) (Y3 - {b13}*Y1 - {b14}*X1 - {b15}*X2 - {b16}*X3 - {b17}*Z2 - {b18}), xtinstruments(X1 X2 X3, lag(0/1)) xtinstruments(2: Z1, lag(0/1)) xtinstruments(3: Z3, lag(0/1)) wmatrix(hac bartlett 1) winitial(xt L) vce(hac bartlett)


I get a new error message that:
"wmatrix(xt ...) inconsistent with number of moment equations"

I have 3 moment equations for the whole panel, and specifying "wmatrix(xt LLL)" gets me the error message:

"Can specify at most two moment equations"

Any help with how to respecify the 3 moment equations will be much appreciated!

Many thanks, May




------------------

-----Original Message-----
From: Kit Baum <kitbaum@me.com>
Sender: owner-statalist@hsphsun2.harvard.edu
Date: Thu, 17 Mar 2011 11:58:13 
To: <statalist@hsphsun2.harvard.edu>
Reply-To: statalist@hsphsun2.harvard.eduSubject: Re:  st: 3SLS with different instruments for different equations

<>
On Mar 17, 2011, at 2:33 AM, May wrote:

> Hi, thanks - I tried gmm specification for my model below by typing:
> 
> . gmm (Y1 - {b1}*Y2 - {b2}*Y3 - {b4}*X1 - {b5}*X2 - {b6}*X3 - {b0}) (Y2 - {b7}*Y1 - {b8}*X1 - {b9}*X2 - {b10}*X3 - {b11}*Z1 - {b12}) (Y3 - {b13}*Y1 - {b14}*X1 - {b15}*X2 - {b16}*X3 - {b17}*Z2 - {b18}), xtinstruments(X1 X2 X3, lag(0/1)) xtinstruments(2: Z1, lag(0/1)) xtinstruments(3: Z3, lag(0/1)) wmatrix(hac bartlett 1)
> 
> But I got an error message:
> 
> "Must specify winitial(xt xtspec) or winitial(matname) with XT-style instruments"
> 
> I tried searching "help winitial" but nothing came up. 
> 
> Can you please help with my gmm syntax?


In a prior posting I pointed out that what you were trying to do made no sense in a 3SLS context due to identification status of the equations. Linear GMM is no magic bullet. As ivregress and ivreg2 illustrate, GMM on a linear model is just a variation on standard least squares techniques. Likewise, linear GMM will not allow you to do something that 3SLS cannot. If a linear equation is not identified, it is not estimable, regardless of the technique used.

Kit Baum
kitbaum@me.com



*
*   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/