First stage of Lewbel(2012) IV method

[kamble vikrant]

Hi All,
I have a question for econometricians on this platform.
I wanted to implement Lewbel's (2012) IV method to panel data.
In this method, instruments are created internally using exogenous variables in the equation.
I was wondering if the instruments yield valid first-stage results after controlling for those exogenous variables.
The ivreg2h command written for implementing this method does not post first-stage results.
Any help to get the first stage results in this method will be appreciated.

Regards,
Vikrant

[George Ford]

This is close, so a good spot to start to fiddle.

The "first" option gives you a ton of stats, so not clear what you're after.

use http://www.stata-press.com/data/r9/grunfeld.dta , clear

ivreg2hF invest L(1/2).kstock (mvalue=), savefirst gen(g,replace) fe first 
ereturn list

center kstock , g(kstock_c)
center mvalue , g(mvalue_c)

reg mvalue gmvalue_L_kstock_g gmvalue_L2_kstock_g L.kstock L2.kstock  , absorb(company)

[George Ford]

You may have to dig into the ado file and create variables from the tempvars (the __C00003 stuff) you can use. I would think the above would reproduce the first stage, but it does not. It's close.

I didn't use the centered variables. It made no difference.

[George Ford]

Also, I used ivreg2hF, which is a rewrite I made trying to extract some information.

I think to figure this out, you'd have go through the ado file to see what each of the relevant parts are doing. I've spent some time with it, but still can't replicate the first stage exactly.