st: unit root in balanced panel

[Andreas Pondorfer <Andreas.Pondorfer@student.uibk.ac.at>]

Dear Statalist,

I have a question concerning unit root test for balanced panels.

I´m examining climate impact on production in Sub-Sahara Africa, therefore
I run a CD-production function with a two way fixed effects model:

yit = a + xitb + vi + Dt + eit

After running the regression, I tested for serrial correlation. The results of
xtserial and xtregar, lbi indicate autocorrelation.

Thereafter I wanted to test if the panel series are stationary. For  
that I used
the ipshin command, which is a test for unit root in heterogenous panels.

Here, I have problems with the interpretation of the unit root test results.
I have seven input variables. Stationarity is given for the output  
variable and for 6 out of 7 input variables. One variable includes a  
unit root.

Do I need to reject stationarity for the whole panel series?

When doing First differences, all the data is stationary.

What do you recommend?

Thanks for your help!





*
*   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/