2
$\begingroup$

I have two stationary time series. I would like to check for cointegration between them. Does this make sense, and can I just use Engle-Granger Test (two step) for Cointegration for this?

$\endgroup$
  • $\begingroup$ I come to think of it that just plain correlation or (adjusted) R^2 are the same for stationary processes? $\endgroup$ – BigChief Aug 20 '14 at 10:34
1
$\begingroup$

No, it does not make sense to look for cointegration among stationary time series. Cointegration can only take place if the individual time series are integrated (thus non-stationary).

The basic idea can be found in Wikipedia: If two or more series are individually integrated ... but some linear combination of them has a lower order of integration, then the series are said to be cointegrated.

$\endgroup$
  • $\begingroup$ already got the answer, one could use just correlation $\endgroup$ – BigChief Feb 11 '15 at 11:46
1
$\begingroup$

Yes that is absolutely true, if the series are already stationary at levels , running a cointegration does not make sense ( It requires data to be I(1) or integrated of the same order). If the data are already stationary then it makes sense to proceed with VAR.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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