[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: re: 2sls robust or 3sls?

From   Kit Baum <>
Subject   st: re: 2sls robust or 3sls?
Date   Sun, 27 Sep 2009 10:20:52 -0400

Andres said

I'm dealing with a simultaneous equation system, three supply and
three demand equations. As endogeneity is present, a instrumental
variable method is required. I have 37 years of historical data.

- I'm in doubt to use 2sls (robust) or 3sls. I know that 3sls is
more efficient than 2sls but because of the small number of
observations it might not be good idea.

- 3sls doesnt allow the option robust. Is there any way to do so?

- when I use reg3 for the whole system together, the results are
different if you I it equation by equation. Anyone know the reason
or which way is more reasonable?

John A. pointed out that if the results (e.g., point estimates) differ, this is likely to be indicative of misspecification in at least one equation. This is an excellent point, and one of the reasons why 3sls is often not such a good idea--especially the crippled form of 3sls available from -reg3-. Unlike almost all other Stata estimation commands, -reg3- does not support a -vce()- option, so that your -reg3- estimates must be calculated under the maintained hypothesis of i.i.d. errors (which is often a ridiculous assumption). In that sense efficiency gains are irrelevant, as your -reg3- standard errors based on i.i.d. errors are likely to be biased and inconsistent.

I would come down strongly in favor of 2sls with robust (or HAC) standard errors and diagnostics.

Kit Baum   |   Boston College Economics & DIW Berlin   |
An Introduction to Stata Programming |
   An Introduction to Modern Econometrics Using Stata  |

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index