Re: st: xtabond2 with static model?

[Christopher Baum <kit.baum@bc.edu>]

On Dec 1, 2010, at 2:33 AM, Agnes wrote:

> I have a panel dataset and want to do a regression were one of my explanatory variables is probably endogenous, more specific there could be a problem of reverse causality.
> Since there are no appropriate instruments, I use a within estimation (fixed effects).

Note that including fixed effects has nothing to do with solving a problem of endogeneity. Fixed effects deal with unobserved heterogeneity at the panel level. They do not and cannot help if the error term is correlated with some of your X variables, even after the within transformation.

I agree with the other suggestion you have received: use xtivreg (or Schaffer's xtivreg2, from SSC, which provides additional diagnostics) and fit the model using lagged values of the potentially endogenous variables. The endog() option in xtivreg2 will allow you to test for that endogeneity, as described in the -ivreg2- help file (ivreg2 from SSC, along with ranktest, must be installed with xtivreg2). Include a lagged dep var if you question whether a static specification of the equation is appropriate.

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


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/