I have revised the Stata program xtabond2 (and reposted to SSC archive) a
number of times since I released the program last November. (This program
does Arellano-Bond and Blundell-Bond dynamic panel estimation.) Most of the
changes added features or fixed bugs that would cause crashes in unusual
situations. However, two recent updates fixed bugs that can lead to wrong
results.
So if you use this program, I *strongly recommend* downloading the latest
version now by typing "ssc install xtabond2, replace" in Stata. And I
apologize for these bugs.
New features:
1. There is now an arlevels option to apply the Arellano-Bond AR() test to
the levels residuals instead of the first-difference residuals-appropriate
only in system estimation.
2. The gmmstyle() declaration now has an equation() suboption that can
suppress the generation of moment conditions for levels equations in system
estimation.
More detail is in the help file.
Bug fixes to prevent wrong results:
1. xtabond2 no longer ignores gmmstyle() the laglimits() suboptions when
the collapse suboption is also invoked. (Thanks to Fabio Schiantarelli.)
2. In system estimation, xtabond2 now generates "GMM-style" moment
conditions for the levels equations even for time periods that have no
complete observations in first differences. (Thanks to Tue Gorgens.)
3. In system estimation, in "IV-style" instruments, xtabond2 sometimes put
first differences of the instruments in the rows for the levels equations.
It no longer does. (Thanks to Tue Gorgens.)
--David
David Roodman
Research Fellow
Center for Global Development
1776 Massachusetts Ave. NW
Suite 301
Washington, DC 20036
voice: +1-202-416-0723
fax: +1-202-416-0750
|