Tuesday, December 3, 2013

Let's disaggregate aggregation

I recently learned of a firm that has enhanced their system to provide the ability to aggregate multiple portfolios for performance reporting. They reference GIPS(R), so clearly the fact that it's permitted within the Standards for composite returns is seen as justification for their work.

When embarking on any performance or risk development / design work, it is extremely important to understand WHAT is being done; the PURPOSE of it; what it is intended TO DO.

I have many times criticized the use of the aggregate method; to me, it's seriously flawed. I've shown examples of the flaws, which apparently prompted the GIPS Executive Committee to declare the ability to add accounts intra-month invalid (which I'm fine with, but would have preferred a public comment period, since this was a change to the Standards, and as such, warranted input from those affected).

Are we looking for (a) the average return of (b) the return of the portfolios, as if they constituted a single account? Chances are the former, which would have dismissed this approach.

There are many ways to calculate returns; it's important to spend the time understanding what is intended in order to properly select the ideal approach.


  1. To run performance attribution analysis on a "composite" and/or strategy, the aggregate approach makes the most sense. For composite reporting purposes, using each accounts' returns and weights provides additional, beneficial information not available if using that aggregate return. It goes back to your comment above, what purpose does this data serve.

  2. Debi, very interesting point about running attribution against it. You haven't completely sold me, but it's worth reflecting on. Glad we agree on the latter point. Thanks!


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