Granger Causality Test. Stationary or non-stationary data? (Eviews)

[Herbert]

Hi,

Is it correct to use Granger Causality Tests on non-stationary, i(1), time-series? Or should I (log)difference my data first?

I cannot seem to find the answer to this question in any of my books or online.

Thank you very much for your help.

[erick]

Dear sir,
I also had the same problem. Hopefully, I got something which can be helpful for both of us.
In sum :You need to have stationary data to run the Granger test.
Please read the following document for more details
http://mpra.ub.uni-muenchen.de/2962/...paper_2962.pdf

You can read on page 5 this : "Granger causality requires that the series have to be covariance stationary......."