Revalue portfolios, not composites, for large cash flows (at least when the flow is a new account!)

Effective 1 January 2010, GIPS(R) (Global Investment Performance Standards) compliant firms are obligated to revalue their portfolios for large cash flows. A Q&A made it clear that this applies to firms that use the "aggregate method" to derive their composite returns, even though the underlying portfolios aren't used in this method to come up with the return. However, I would argue that it would be inappropriate to revalue composites themselves for large flows! I first addressed this in our November 2006 newsletter, following a project I conducted for a unit of Bear Stearns (I'm comfortable mentioning the client now, given their demise). The project involved a case where they grouped accounts together to provide a client with an overall representation of their performance. And while I later argued for the use of money-weighting in these cases (not for GIPS purposes, but for client reporting), I nevertheless tackled the time-weighting angle. It seems that they replaced a prior method with one that revalued the collection (aka "composite") when large flows occurred. At first, I thought this was probably acceptable; however, the results proved nonsensical, which caused me to reflect more about this.

I raise this issue again because (a) it has been some time since I first tackled this, (b) GIPS now requires revaluation for large flows, and (c) a (new) client presented a problem they were having with returns which smacked of this exact issue, and so I will repeat myself a bit, so as to reaffirm the harm that can result from revaluing composites for large flows.

Here's the scenario as laid out in the November 2006 newsletter:

Pretty simple case, right? We have a composite that begins with a single account, which is joined half way through the month by a second account. Should we revalue the composite because of the large flow?

Well, if we do the result will be 1.33 percent. Does this make any sense? After all, the two portfolios have returns that both exceed this value, so how can this be right? When I was presented with an example like this from Bear, I was at first perplexed, because I couldn't see why the result would be so obviously wrong!

If, however, we use an asset-weighted approach that avoids revaluation, the result is the more plausible 3.33 percent! I will leave it to the reader to either validate these results on your own or to visit the earlier newsletter which provides the details.

And so, how can we explain what's happening? Let's revisit the reason we revalue: to eliminate the impact of the cash flow on the portfolio, so as to separate the timing before and after the flow event. However, is the manager managing the composite? Hardly! They're managing two separate portfolios. To revalue the first portfolio because a new one is introduced causes us to capture a point where the account had dropped in value: but what does this have to do with the second portfolio being added? Nothing!

Firms that go to the trouble of revaluing composites for large flows are opening themselves to reporting erroneous results when they revalue for new accounts being added during the month. Granted, most firms bring on new accounts at the start of a new month, so this wouldn't be an issue, but for those that will bring an account in mid-period, problems can arise and arguably any result is an error.

Should the firm revalue for flows other than for new accounts? I'm not sure about this as it would require more thought on my part. But for now, let's say I'm uncomfortable with the notion of revaluing an entire composite because one of the accounts has a very large flow: I don't believe the result will necessarily be accurate.

p.s., I want to point out that the result I obtained without revaluing was by using the asset-weighted method, not the aggregate method. I am exploring the impact of revaluing vs. not on the aggregate method and will post something shortly.

p.p.s., I also want to emphasize that my current position about revaluing is regarding cases where new accounts are added to a composite during the period; I am researching the broader case of revaluing.