I have multiple variables that I am trying to perform a VAR model with, but all my variables are non-stationary at levels as they fail the Augmented Dickey Fuller test. Having taken the first difference, they do all become stationary at I(1). I was wondering, do I perform the Johansen cointegration test on the I(0) variables or the I(1) variables.

Thanks in advance

(I also have a couple other questions, please do answer this as well if you know the answer: 1) I am right in thinking I should perform the VAR regressions on the I(1) variables, right? 2) If I have cointegration present, and my I(0) variables are non-stationary, I should switch to a VECM model? )


1 Answer 1


You do the cointegration analysis/test on I(1) variables, because no cointegration is possible among I(0) variables. (This is textbook material; try some of these.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.