Dear all,

I have been doing some reading on the importance of the stationarity property for the dependent and independent variables when doing OLS regressions. On some forums, I have read that if one of the regressors, dependent variable, or the error term has a unit root, the OLS estimator is no longer BLUE and that you cannot use reg/xtreg. Whereas on other forums, I have read that for panels this is not a problem, especially in cases where the N is larger than T. I am now not sure which answer (if any) is true, and was wondering if any of you is familiar with this? If this is indeed a problem, any suggestions what model can be used alternatively then? Thank you in advance. Your advice is greatly appreciated.

Best,

Satya