Dear Statalisters,
I have a problem concerning the right choice of command/options in a panel
data setting. I am working on a macroeconomic panel of ca. 80 countries over
10-25 years per country, i.e. unbalanced. Random effects are rejected in
favour of fixed effects. Furthermore, I have identified both group-wise
heteroskedasticity using -xttest3- (I know power is low, but the test given
in Stata's FAQs using -lrtst- gives a similar result) and contemporaneous
cross-section correlation using -xttest2-.
Now, for fixed effects I could use -xtreg, fe-, but this does not allow
corrections for heteroskedasticity and the correlation. On the other hand,
-xtpcse- corrects for all this, but uses OLS, thus does not have fixed
effects. Indeed, some coefficients differ quite a lot comparing the results
using the two estimators.
It now seems to me as if I only have the choice between having biased
coefficients (using -xtpcse-) or wrong standard errors (using -xtreg, fe-).
As a last solution, I could try to implement an LSDV approach, using
-xtpcse- with mean-differenced data or with individual dummies. Is this
correct? Or does anybody know of an easier procedure provided by Stata to
deal with this? It would be very helpful!
Markus Poschke
Maastricht University
*
* 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/
