15
$\begingroup$

There are several methods available between data vendors and associated software programs to adjust futures contract data for historical simulations.

Some of the methods are:

1) Back or forward Adjust by iterative level shifting of historical data at rollover points.

2) Back or forward Adjust by iteratively using ratio of historical series to current series.

3) Using next contract with highest volume instead of front contract as the next contract to be joined.

Some of the problems associated with different methods can be

1) negative price series

2) inconsistency in rollover dates (one person may use first notice day e.g. commodities, another might use 20 days prior to settlement, another 7 etc..)

3) adjustment methods vary by instrument.

4) There could be true price gaps between the rollover period that have been filtered out by the method of adjustment.

The list is nowhere near exhaustive* Could anyone elaborate on what methods you or your firm are actually using to systematically adjust a broad portfolio of futures contracts (for simulations) in practice? Also, what pros and cons are there to using your method vs. some of the other methods that exist? Any references to expand on your method are also appreciated.

*There are some papers e.g. here and here and blogs that discuss pros and cons, but I'm more interested in best practices and approaches by large portfolio managers.

$\endgroup$

3 Answers 3

7
$\begingroup$

As you pointed out there are many ways to adjust for the roll overs. Hence, I guess you would agree that there is no one-size-fits-all answer to this. It really depends on the usage of the data:

  • First think about how the trades in your back test are structured. If they are longer-term trades and you hold over roll overs then think what you would do if you traded the contract. Meaning, would you try to trade the basis? Would you simply close out and open a day before expiration? Whatever you intend to do you also then need to apply to the futures roll over in your back test. Thats why I said there is no general answer that fits it all. But here couple other points to consider:

  • If you strictly peruse the data for intraday analysis/back tests then you do not and probably should not make any adjustments to the data itself.

  • If you need to create a "continuous contract" (I just use the terminology loosely) then I would generally backward-adjust which is what most reputable data vendors do. Though I also have worked with un-adjusted data and simply adjusted, ratio-wise, on the fly. I do not think there is a strict right or wrong answer. But if must make up my mind I would say, backward adjust.

  • Regarding trading I would roll over to trade the next contract when volume in the new contract exceeds the one of the old contract. I would potentially not trade the current month beyond first-notice in case I am subject to receiving notice.

  • I would potentially skip the next contract if a contract another 1-2 months out trades at much more volume. Obviously this does not apply if your strategy relies on the specific contract month (for example as hedge in an options book or because you trade basis or there could be many other reasons).

$\endgroup$
2
  • $\begingroup$ What do you think about using optimization to drive the transactions as compared to rules like these? $\endgroup$
    – John
    Feb 8, 2013 at 1:35
  • $\begingroup$ @John, by all means why not. It depends of course whether you look to develop a new strategy that benefits from such optimizations. I am well aware that buy side funds or certain ETFs spend an awful amount of time to optimize futures roll overs. The above are more like general guidelines rather than a must-do. $\endgroup$
    – Matt Wolf
    Feb 8, 2013 at 2:34
4
$\begingroup$

An oldie but goodie from 2000. Bob Fulks, Back-Adjusting Futures Contracts, http://www.nuclearphynance.com/User%20Files/7228/cntcontr.pdf

The article is a great summary of the most popular adjustment methods. I agree with and would like to stress the author's observation: "There is no "best" method in an absolute sense. All the methods have advantages and disadvantages." Also, IMHO, some of the original writings on continuous adjustment by Jack D. Schwager are worth investigating.

$\endgroup$
3
  • $\begingroup$ Hi SpzToid, welcome to Quant.SE! Can you elaborate on this paper a bit, why do you think it's good? $\endgroup$
    – Bob Jansen
    Aug 29, 2014 at 8:50
  • 1
    $\begingroup$ The article is a great summary of the most popular adjustment methods. I agree with and would like to stress the author's observation: "There is no "best" method in an absolute sense. All the methods have advantages and disadvantages." Also, IMHO, some of the original writings on continuous adjustment by Jack D. Schwager are also worth investigating. $\endgroup$
    – SpzToid
    Sep 2, 2014 at 8:15
  • 1
    $\begingroup$ Thank you for your clarification. I've added the comment to your answer to give it more prominence. $\endgroup$
    – Bob Jansen
    Sep 2, 2014 at 9:01
2
$\begingroup$

This article has some interesting descriptions about the different methods used and when you would them: https://www.quandl.com/collections/futures/continuous

Thought it might be somewhat relevant to your question.

$\endgroup$

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.