re: st: xttest0 confusion

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

<>
Thank you Kit for your reply. I have done the Hausman test and Wooldridge's robust Hausamn test, both suggest that RE is consistent. The confusion is if the country effects are random (as identified by BP test), still there is serial correlation in the errors. What is the econometric reason for it? If it is because of endogeneity, then why BP is saying the country effects are random and we assume that the eit is iid?
 

Serial correlation in the errors in any time-series regression context is often an indication of misspecification of the equation: that is, there are omitted variables, misspecified dynamics, omitted lags of included variables, omitted interactions of included variables or inadequate functional forms of included variables. The maintained hypothesis for the Hausman test is that the fe model is consistent, regardless of the orthogonality of regressor and error. If that is violated, then neither fe nor re will be consistent. I would try a more general specification (including, e.g., some powers and interactions of your regressors) and see whether those additional terms can be properly excluded from the model. 

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/