Wednesday, March 7, 2012

What were they thinking?

Those who were around "at the creation" recall the debates regarding whether composite returns should be equal- or asset-weighted. Two groups in particular, the ICAA (Investment Council Association of America; now the IAA) and IMCA (Investment Management  Consultants' Association), lobbied AIMR (Association for Investment Management and Research; what is now the CFA Institute) for equal-weighting. I'll confess that at the time, I didn't pay this a whole lot of attention, and didn't formulate an opinion.

AIMR wanted the composite return to represent the experience of a "single account." That is, what the return would be if the composite was an account itself. IMCA and the ICAA felt that asset-weighting might influence some managers to favor larger accounts, knowing  that their returns would skew the results. And, I suspect that they also thought that equal-weighting made more sense as it shows the average return of actual accounts. But AIMR was steadfast ("resolute," in "W" speak) in their position, and refused to budge. IMCA was so determined that they created their own standard, which went into effect the same time the AIMR-PPS(R) did: it never caught on, however.

The AIMR-PPS did, of course, catch on, and motivated other countries to develop standards, which led to the creation of the Global Investment Performance Standards (GIPS(R)). And as with the AIMR-PPS, asset-weighting because the required way to derive composite returns.

But why? What is the benefit of the composite looking like an account, when it isn't one? The composite is comprised of one or more real accounts, that were managed individually; no one "managed" the composite. Would it not be better to see the average experience of real accounts?

When I conduct GIPS verifications I occassionally run across cases that SHOUT OUT to me that this is all wrong. Here's one recent example:

Because of the huge size difference, account A's return IS the composite's: account B doesn't even have to show up. What's the point of worrying about B? It has zero influence on the return. And yet, the manager's ACTUAL performance in this discipline lies between these two accounts: actually RIGHT IN THE MIDDLE of them (what mathematicians and statisticians call, the average)!

Okay, so the Standards recommend that firms show the equal-weighted composite return. Great! How many firms do? The number is approximately zero. And why not? Perhaps it's because they would prefer not to hand out their presentations on legal size (i.e., 8 1/2" x 14") paper, or resort to a 9 or 10 point font size to fit everything that's required on the page.

I know that this commentary is about as welcome to some as ants at a picnic. But seriously, what were they thinking when they advocated asset-weighting? NO ONE MANAGES COMPOSITES! Firms don't get paid TO MANAGE A COMPOSITE! Would it really be so bad to say, "okay, maybe equal-weighting makes more sense, so effective 1 January 2015, equal-weighting will be mandatory, asset-weighting is optional, and the change goes into effect on this date, but firms are encouraged to restate history"? And what's the likelihood of this occurring? Again, approximately zero. Oh, well.

p.s., Yes, the figures in the table come from a client, though they've been altered slightly, out of respect for our client's confidentiality.


  1. David,
    I believe that an important argument against equal weighting is its lack of continuity.
    By the requirement of ‘continuity’ I mean that every evaluative measure must satisfy the following criteria: A negligibly small change in any of its continuous inputs must create a negligible change in the result.
    Equal weighting does not satisfy this requirement. The value of an equal-weighted return of a $100B composite of two accounts of significantly different returns would change discontinuously if the value of its smallest account went from $1 to zero dollars.
    Equal weighting is unacceptable because it allows negligibly small changes in the market value of one of its funds to have a non-negligible impact on the resulting equal-weighted return.
    To address some of your other comments: There are many different kinds of averages and I believe that a PM’s responsibility is properly more focused on where more of the money entrusted to her is.

  2. Andre, while I appreciate you taking the time to comment, your argument is fundamentally flawed, because you're speaking to the problem that might be inherent in a return itself, which is an entirely different matter. Yes, if the return of a single account went from $1 to zero dollars, we'd have a problem, but this would also be a situation where the account doesn't belong in the composite in the first place! There is an expectation and understanding that accounts in composites (a) belong there, (b) have returns that were calcualted properly, and (c) that the returns, individually and collectively, represent the performance of the stated strategy. Your argument against equal-weighting is based on a flawed assumption, and therefore is flawed itself. Sorry.

  3. Dave,

    I understood the primary advantage to size-weighed composites is that they cannot be gamed. Often, investment advisors have limited access to some exceptional investment opportunities. For example, in the case of popular (oversubscribed) IPOs the advisor is allocated the in-demand shares based on the amount of trading they do. It is possible (though not ethical or necessarily legal) to assign all of the shares to small portfolios. Those portfolios would then raise the return on an equally weighted composite.

    I believe that if there are two methodologies available for a set of performance standards, and one can be manipulated while the other cannot be manipulated, the latter is preferable to the former.

  4. Neil, thanks for your note. Kind of funny how you are taking the opposite view of what the ICAA & IMCA did, suggesting that in a size-weighted approach, a manager would favor the larger clients, because their higher returns would skew the results. In both arguments (yours and theirs), the belief is that the manager will "game" the returns. No doubt this can happen, and no doubt it can happen in both instances. And, in both cases, there should be checks done (e.g., when the SEC shows up) to detect such behavior. And so, I think it is fair and accurate to say, that for asset-weighted returns, the ICAA & IMCA's argument is that the manager can "game" the composite results, by favoring the larger accounts that will skew the returns; and, in equal-weigting and your argument, the manager can "game" the returns by favoring the smaller accounts.

    And so, if we put the "gaming" aside, and simply consider which is better, to me equal-weighting is, because it shows how, on (true) average, the manager did across all accounts, irrespective of size.

  5. Stephen Campisi, Client-Centered InvestorMarch 8, 2012 at 11:42 PM

    Excellent points on both sides, although none of them address the essential issue. (Once again we have a case of expert practitioners arguing over methodology when they have not agreed on the question to be answered.) So, let's first address the purpose of the performance standards: to provide a consistent and credible representation of a manager's performance record for the set of client portfolios managed in a specific style or mandate. Now, let's look at the question that the potential client (AKA the prospect) wants to have answered: "What is the return that I would have achieved if I had invested with this manager?" More to the point, the client wants to know the average CLIENT return rather than the return on the average dollar invested. Put bluntly: "What's MY likelihood of getting the composite return?" With a size weighted composite, the answer is "Less likely than if we showed you an equal weighted composite return."

    It seems prudent to have some healthy skepticism regarding the potential for larger clients receiving preferential treatment and thus earning higher returns; in this case a size-weighted composite inflates the return. But again, the question we are responsible for answering to satisfy our duty to the prospect requires an equal weighted average return for the composite.

    None of you have addressed the issue of risk. Bear in mind that a size weighted return approach will bring with it a downward biased standard deviation risk statistic. Taking that in combination with a potentially upward-biased return and you have the potential for a significantly biased risk adjusted return being shown to potential clients. I believe that an equal-weighted approach is more conservative and finds yet a third reason why it is preferable from the standpoint of the client, whom the performance standards are supposed to serve.


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