Futures Correlations

Discussion in 'Technical Analysis' started by maninjapan, Dec 1, 2009.

  1. I would like to check the correlations of some futures contracts but not sure if I need to adjust the contracts or not. Do I just run it based on the price? Or do I need to change it to a dollar value (close multiplied by point value) to get a true correlation?

    Any comments on this would be much appreciated.

  2. If you're working with "linear correlation" (as opposed to say co-integration), and if you want to be able to compare your calculated correlation coefficients with anybody else's, just make sure that whichever of price or points you choose to work with, you do the same for both futures contracts in each pair.

    Apologies if I have misunderstood your question in any way ...
  3. Sorry, what do you mean by price point?
  4. mah56j


    if you do this without a continuus contract, consider doing the correlation contract by contract if they two series are "co-terminus". that is, if they expire on the same date. I would try to avoid low volume data points.

    if using cont contracts, I would still use multiple time samples. Some correlations will cycle over time.
  5. MTE


    You should use returns to estimate correlation and not the actual price.
  6. guys, thanks for the replies. I plan to use continous contracts. Just wasnt sure wether I needed to do it jsut based on price, or absolute returns.

    MTE, thansk for that looks like I need to multiply by the point value.

    Thanks guys!
  7. MTE


    You don't need the point value. You just calculate the daily (or whichever period you like) returns from the price of each contract and then run the correlation.

    The absolute value is completely irrelevant.
  8. If there is an underline index to those future contracts, better use that. Some futures, like crude oil, show negative prices in the distant past after many adjustements. Be aware of that.
  9. sorry MTE, guess I dont quite get it. whats the difference between just the price and the absolute return?
    More importantly, how do I prepare the data before I run a correlation calculation?

    Thanks for your help
    #10     Dec 1, 2009