Wednesday, July 25, 2012

What do you need to do when you change your return methodology?

Last week I began to conduct a software certification for a Spaulding Group software vendor client. My primary contact mentioned that they changed their policy for timing of cash flows, so that inflows default to start-of-day events, while outflows default to end-of-day (though these can be overridden, if necessary). She asked if anything needs to be disclosed, and also, what happens if they recalculate history because of an as-of trade?

First, we would not expect anyone to recalculate history, just because a change is made to the formula, with only one exception: if the method was flawed (we've found flawed formulas used at a few firms, who developed their own system, using their own in-house designed formula). Okay, maybe a second example: if an asset manager wants to comply with GIPS(R) (Global Investment Performance Standards), and the one(s) they had been using historically don't comply with GIPS, so in order to comply, they have to recalculate past performance).

As for disclosures, I'd say that they need to indicate the timing of changes and what formula(s) is(are) used at which time.

This disclosure can also include a statement that in the event of as-of trading history is recalculated, the newest employed formula at that point in time will be used.

Make sense? Have a different view? Chime in!

4 comments:

  1. David,

    Since you mention GIPS, I presume that this note is focused upon Fund-Level returns and is not meant to address the much more difficult problem of calculating the returns of components of funds.
    Also, lets put aside any methodology that allows the user to freely decide how its gets employed on a case-by-case basis.
    How do you judge whether a (return) methodology is “Flawed”?
    I would think that to designate a model as flawed enough to disallow its use, it is sufficient to point out that there are cases where it leads to absurdity.
    Do you have another criterion? A criterion that might permit models that my suggestion would not?
    That is, are there cases where you are willing to designate a model as not flawed, and thus acceptable for use, even though there might be certain situations where that model gives unacceptable results?
    Or do you agree that if a model ever gives unacceptable results then that model should always be disallowed?

    Andre

    ReplyDelete
  2. Andre, thanks for your comments. I am limiting this to any particular level of performance, as I think that the "rules" should apply, regardless. If you are making a change to a method, where the prior method produced what would be considered acceptable results, and your change does, as well, then other than documenting the change, not much more is needed; even GIPS doesn't require a disclosure.

    As for what constitutes a "flawed" method, I would say any method that produces erroneous or misleading results. I am conducting a review of a client's system this week, and their approach is totally unique (I won't go into the details; sufficite to say, it's probably like no other you've seen before); that being said, it produces reasonable results. It is not flawed. I have encountered methods that are flawed, in that they are built upon nothing other than the creativity of the designers, who ignored the vast amount of literature on the subject. I will post something along htis lines shortly.

    ReplyDelete
  3. Dave,

    I have to say I am against software houses dictating calculation methods and deciding that they will simply choose another (albeit “better”) method. I would prefer that a system setting is introduced so that clients can determine whether they wish to stay with the old method, or adopt the new one. Even better, the methodology could be time-stamped to allow a change over time (disclosed, of course). This provides for a more flexible solution – better for the client and, ultimately, better for the vendor as the system ticks more RFP boxes.

    An extreme example would be a vendor who decides that their composite solution should use asset-weighted rather than equal weighted dispersion (as I know you are a fan of the latter!). Clearly GIPS allows this but I doubt that managers using the system would be happy to change.

    All the best

    Anthony

    ReplyDelete
  4. Anthony, interesting points. I agree that the vendor should have flexibility in their system to allow clients to continue the old way, if they prefer. One of our clients replaced their prior money-weighted formula (Modified Dietz) with the IRR, which is much more accurate, but continued to allow their clients to continue w/MD, if they wanted. Likewise, the vendor client of ours who moved to Start of Day for inflows and End of Day for outflows allows their clients to continue with their prior method. I think this should be a standard approach, unless (of course) the prior method was determined to be flawed.

    ReplyDelete

Note: Only a member of this blog may post a comment.