FYI The following message appeared on the Econometric-Research list but
does not appear to have been posted on Statalist. I'm taking the
liberty of forwarding it, given the likely interest in it. (The command
mentioned is available via -ssc install xtabond2-
Stephen
-------------------------------------------------------------
Professor Stephen P. Jenkins <[email protected]>
Institute for Social and Economic Research
University of Essex, Colchester CO4 3SQ, U.K.
Tel: +44 1206 873374. Fax: +44 1206 873151.
http://www.iser.essex.ac.uk
-----Original Message-----
From: A broadcast forum for applied and theoretical researchers in
econometrics. [mailto:[email protected]] On Behalf Of
David Roodman
Sent: 26 November 2003 17:51
To: [email protected]
Subject: new Stata command to do Arellano-Bond/Blundell Bover "system
GMM"
I'm pleased to announce a new Stata command, xtabond2, that implements
the "system GMM" estimator, first outlined by Arellano and Bover in 1995
and fully developed by Blundell and Bond in 1998. xtabond2 can also do
the original Arellano-Bond "difference GMM" estimator, much like Stata's
official xtabond. xtabond2 differs from xtabond in that:
1) It can do system GMM
2) Its syntax is quite different
3) For both difference and system GMM, it makes available the Windmeijer
(2000) finite-sample correction to the two-step covariance matrix. This
makes two-step estimation a viable alternative to one-step, robust.
4) For one-step, robust estimation, it reports the Hansen J test of
overidentifying restrictions, which is actually the standard Sargan test
for the two-step estimator. Unlike the Sargan test for the one-step
estimator, the Hansen J is robust to heteroskedasticity and
autocorrelation within panels.
5) It gives the user more control over the instrument matrix
6) It accepts time series operators in all variable arguments
7) It's slow (Any enterprising C programmer out there want to help me
with
this?)
xtabond2 works with Stata 7.0 and later.
You can install it by typing "net install xtabond2" in Stata. Then to
learn more about it, type "help xtabond2".
Also available is a .do file that exactly replicates the results of a
sample program distributed with the DPD for Ox, the original
implementation of system GMM. To obtain it, type "net get xtabond2".
Note that this file uses an undocumented xtabond2 option, dpds2, to
replicate what I believe is bug in DPD for Ox that affects the AR(i)
tests in one-step system GMM. The DPD output for comparison is the file
"bbest1.out" in
http://www.nuff.ox.ac.uk/Users/Doornik/software/dpdox121.zip.
I welcome comments and suggestions.
David Roodman
Research Fellow
Center for Global Development
1776 Massachusetts Ave. NW
Suite 301
Washington, DC 20036
[email protected]
voice: +1-202-416-0723
fax: +1-413-410-2602
*
* 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/
