[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: re: overidentification

From   Kit Baum <[email protected]>
To   [email protected]
Subject   st: re: overidentification
Date   Fri, 7 Sep 2007 09:37:22 -0400

Anna wrote

My dependent variable is education and my key explanatory variable is
institutions which I have instrumented using Acemoglu's settler mortality
rates. I have ran 2sls regressions with alternative instruments instead of
settler mortality rates but then I cannot work out how to get stata to work out
p-vale for a chi squared test to test whether the 2SLS coefficients (estimated
with the alternative instruments) versus the coefficients estimated using
(log)settler mortality (the IV i have used in this dissertation) in addition to
the “true” instruments are significantly different.

(I guess that the null hypthesis for the p value would be that the coefficient
on average protection against expropriation risk (the institutions variable) in
the second stage regression is the same as when instrumented using log mortality
of settlers in addition to the alternative instruments)

I have been advised to use ivreg2 however I am not sure exactly what commands I
need or how to write the commands?

Please see BC WP 667 by Baum/Schaffer/Stillman on the URL below. It was released since I suggested the use of -ivreg2-. It discusses the kinds of tests you need to do extensively. Better yet, as two of the three coauthors of -ivreg2- and of this paper will be at Cass Business School next Monday and Tuesday, and I will be giving an hour's lecture on instrumental variables Tuesday afternoon

do try to drop by and we can chat about the particulars of your problem.


Kit Baum, Boston College Economics and DIW Berlin
An Introduction to Modern Econometrics Using Stata:

* For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index