Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: xtivreg


From   Kit Baum <baum@bc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: xtivreg
Date   Wed, 31 Mar 2004 10:48:24 -0500

On Mar 31, 2004, at 2:33 AM, Giorgio wrote:

is it correct? how does it work? Are you sure that I've been instrumenting GDP with L.GDP and POP with L.POP?
As has been oft mentioned on this list, this is not how instrumental variables works. You estimate y=Xb+e with a set of instruments Z. Some of the elements in Z are likely also in X (at least a units vector in most cases). There is no meaningful concept of 'this is an instrument for that'--the only things that matter are the order condition (which is a relation on the number of columns of Z vs. X) and the rank condition (that states that these columns of Z are not linearly dependent). In the example you give

xtivreg trade rer (gdp pop = L.gdp L.pop),fe

the equation is exactly ID, in that you have just enough instruments to identify the RHS two endogenous variables. Do you really believe that over the timespan of your data that the country's population is determined by trade? I can imagine a long-run relation where pop.growth is related to openness or lack thereof (consider all those starving North Koreans) but it is a bit implausible to consider that trade(i,t) determines pop(i.t). For one thing, there is a nine month lag involved, so if you haven't conceived by tonight, your child will not be appearing in the 2004 census.

Kit

*
* 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/




© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index