Thursday, February 18, 2010

Standard deviaton: equal- or asset-weight

Let's address the subject of standard deviation, but not from a risk perspective, but rather as a dispersion measure.

Way back in 1997, when the AIMR Performance Presentation Standards (AIMR-PPS(R)) introduced the requirement for firms to disclose a measure of dispersion, they encouraged firms to show asset-weighted standard deviation rather than equal-weighted, because, after all, the composite returns were asset-weighted, why shouldn't dispersion? AIMR introduced a formula to do just this.

Well, a funny thing happened when we went to the last edition of the Global Investment Performance Standards (GIPS(R)) in 2006: nowhere do we find the asset-weighted standard deviation. Where did it go? The Handbook provides the math to derive the equal-weighted measure but not asset-weighted. Other than in Q&As, we see nary a word on asset-weighting. Is this a sign of a change in belief in the value of the asset-weighted approach? I believe it is, although firms can continue to show asset-weighted, if they would like.

But why would you? How do you interpret or explain it? Equal-weighted standard deviation has an understood meaning, but asset-weighting, to my knowledge, doesn't. It's more complex to derive and provides you with what benefit? And, why would using the beginning of the year market value for an annual measure improve upon the tried-and-true equal weighted standard deviation? I say, put an end to this measure and go with the traditional one. It's easier to calculate, is more widely accepted, and is interpretable!


  1. Stephen Campisi, Intuitive Performance SolutionsFebruary 18, 2010 at 12:38 PM

    I believe there is an even better reason to move from asset weighted to equal weighted standard deviation: it's a better answer to the right question. Client's ask about dispersion because their question is this: "How likely am I to receive the average return of the composite?" Obviously, if dispersion among the portfolios in the composite is high, then there is a low likelihood that a new account will earn the return of the composite. So the dispersion question is really this: "What was the dispersion around the average ACCOUNT in the composite?" and not "What was the dispersion around the average DOLLAR (Euro, Pound, etc.) in the composite?" It's clear that asset weighting reduces the dispersion in the composite and so it is not a conservative measure of dispersion risk. It favors the manager by minimizing dispersion.

    Frankly, this argument has also been used for equally weighting RETURN in the composite; alas, this has not met with overwhelming support, as it should have. Perhaps when we get a better understanding of the client's relevant questions, then we will have clarity and consensus around the right measures of risk and return.

  2. Excellent points, thanks! Yes, if someone is looking at the dispersion number, they want to be able go properly assess it: what possible interpretation or meaning can be associated with an asset weighted version? As for equal weighted composites, both the ICAA and IMCA opposed asset-weighting as it gives greater emphasis to larger clients. Because (at the time) AIMR wanted the number to reflect the performance of a single account, asset-weighting ruled and it will never get put back into the bottle...sorry. But, firms can always show equal-weighted composite returns, too!

  3. I also agree, but I believe the core problem rest on the client and not the data provider. If a client wants something, the data provider can't say no. If a client needs something, the data provider still can't say no either. Educating clients to help them understand what the data means is a possible solution, but I feel this always doesn't work the way we expect.


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