st: panel unit root test t=4 / pooled estimation

[Daniel Becker <daniel.becker@uni-rostock.de>]

Dear statalister,

maybe the question is a bit silly, but maybe someone has the time...

I have a panel with 74 cross-sectional observations and 4 year  
(1985,1990,1995,2000). I decided to ignore the time dimension and did  
a regression that includes dummies for the last three out of the four  
years.

However, "regress depvar year" and a scatter plot suggest that my  
dependent variable has a trend. So I thought that I should have a look  
at stationarity. But is this necessary if I am not making use of the  
time dimension anyway?

Secondly, while exploring several tests for unit roots, that most of  
time seem to reject the existence of a unit root, I started to think  
whether those test really are meaningful in my case. In the end, there  
are only 4 years in the panel. (And, in addition, I do not really know  
what an approbiate lag-length might be.)

So my questions are:

Is stationarity important if the estimation is done in a pooled panel?

Are panel unit-root tests valid if the time-dimension is very short?

Thanks for your help,

Daniel




*
*   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/