Statalist


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

st: Re: opinion on 2sls vs 3sls


From   Kit Baum <baum@bc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: opinion on 2sls vs 3sls
Date   Tue, 14 Aug 2007 23:13:22 -0400

One practical consideration, having nothing to do with the econometric techniques but with their implementation in Stata, is that -reg3- is only capable of handling i.i.d. errors. 2SLS techniques, such as -ivreg2- or -ivregress- in Stata 10, can handle violations of i.i.d., including heteroskedasicity, serial correlation and cluster-robust VCEs. On that basis (and the irrelevance of i.i.d. errors to most empirical work) I would much rather use limited- information techniques (2SLS, LIML, or IV-GMM) than 3SLS.


Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


On Aug 14, 2007, at 2:33 AM, statalist-digest wrote:


If you suspect the contemporaneous correlation of
disturbances across your
equations go for 3SLS. Furthermore, assuming that your
system is correctly
specified, 3SLS is more efficient (or at least as
efficient) than 2SLS.

I personally prefer 3SLS. You can also try to estimate
both and to check
the efficiency of your estimates. However, be careful
with your model
specification. If one of the equations is
misspecified, 3SLS may be
inferior to 2SLS.
*
*   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