For example I want to construct 2 time series, one for ES and the other for NQ and test for cointegration.

ES one point equal to 50$.

NQ one point equal to 20$.

If I have the following data:



How do I normalize this data for cointegration test?

TX in advance!


3 Answers 3


Multiply each price series by its multiplier to get notional values. Then proceed as if the notional value were the price of 1 share.


In a co-integration test you rely on the original price series -- not transformations of the price series such as rate of change and so on. Seems to me there is no need for normalization.

  • 1
    $\begingroup$ A "normalization" can be a useful nonlinear transformation quite different from differencing (which should be generally avoided, as you suggest - unless one is dealing with higher order cointegration). $\endgroup$
    – Ryogi
    Oct 27, 2011 at 22:08
  • $\begingroup$ excellent point $\endgroup$ Oct 28, 2011 at 4:23
  • $\begingroup$ but my problem quantGuy is that if you want to trade such a portfolio the different leverage of each instrument is not taken into account.What I think should be done is multiply the price by the point value and then look for cointegration. Does it make sense to you? $\endgroup$
    – Freewind
    Oct 28, 2011 at 8:39
  • $\begingroup$ I see. Yes, that makes sense. Your assumption is that the notional value per @user508's point is stationary. You test for cointegration after adjusting for leverage. $\endgroup$ Oct 28, 2011 at 15:28

It sounds like you want to look at constant volatility series. Just divide the returns by some volatility measure on each series respectively (e.g. ($r_t * volatilitytarget)/\sigma(50 days)$). Then the change in one will be equivalent in size to the change in the other.


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.