Specifically, I'm looking at 3 different commodity futures prices. I want to test that each is unit root and that they are cointegrated. Next, I'd like to apply a model to them to test if they (or a combination of them) are mean reverting.

What I want to see is the evolution of a margin of 3 prices. X=price 1 Y=price 2 Z=price 3 M=Margin

M = (X+Y)*a - Z

first I was going to ADF to test each series for unit roots.

Is it possible to test M directly for mean reversion? I'd like to be able to see when M is too high/low from historic margins and price. Could someone please help??


1 Answer 1


You could as an informal test run a unit root test on your $M$ series.

More formally, it depends on whether $a$ is estimated from the data or chosen by you based on external theory/information:

  • If $a$ it is estimated form the data, then $M$ is measured with imprecision, and hence the test should take this into account as finite-sample correction. This is what the Philips Ouliaris cointegration test actually does.

  • On the other side, if $a$ is chosen by you, you can run the test directly.


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.