Re: st: heteroscedasticity in dymamic fixed effects (or LSDV) models

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

<>
On Jun 10, 2012, at 8:33 AM, Joales wrote:

> 
> I have a usual panel data set on some variables that evolve across
> time and countries.
> 
> I set up a dynamic panel fixed effects model
> 
> and I test for heteroscedasticity using xttest3. So my next step is to use
> 
> xtreg  dependent variable, lagged dependent, rest variables , fe cluster(id)
> 
> where id is the usual panel identifier.
> Checking for heteroscedasticity (using again xttest)  I still find
> that I have heteroscedasticity.
> 
> IS there any problem in my approach?

Yes. It is called Nickell bias (Econometrica, 1981). That is why we have the Anderson-Hsiao and Arellano-Bond/Blundell-Bond estimators.
Computing robust standard errors will generally deal with heteroskedasticiy in this context, and is commonly applied. But you should not apply the fixed effects (LSDV) estimator in this context without computing the potential damage done by Nickell bias -- or better yet, apply an estimator without this bias such as those mentioned above.

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/