Tuesday, November 26, 2013

Annualizing performance attribution

I was in Australia (both Sydney and Melbourne) last week, (1) speaking at a conference, (2) conducting a conference workshop, and (3) teaching our two-day attribution class. It was a very hectic time, and I failed to do any posting, which is a bit unusual for me.

Yesterday, we held our monthly "Think Tank" session, and one caller asked about "annualizing attribution." I found this to be quite an intriguing topic.

It occurred to me that there are three issues that must be considered:
  1. Should we even be reporting attribution for periods greater than a year?
  2. If yes, should we annualize?
  3. If yes, how (i.e., what are the mechanics)?
At the present time, I'm not sure that it's advisable to extend attribution past a year, but need to give this some more thought, as well as discuss the topic with a few colleagues. My concern is that longer period results might "smooth out" the details and provide little benefit to the recipient. However, they may also prove to be quite insightful! In addition to chatting, I will most likely conduct some testing, too, to discover what we might learn.

In the mean time, if you have any thoughts, please let me know.


  1. Stephen Campisi, CFANovember 27, 2013 at 11:29 PM

    Glad to see you posting again. I believe that reporting attribution for periods greater than a year is highly desirable. After all, every active manager's premise is that their clients should expect long term outperformance relative to the passive benchmark. This claim should be substantiated with BOTH the active return differential AND the explanation of the sources of that excess return. In our shop, our active process involves tactical shifts in market exposure as well as active investment selection. We present the long term performance record relative to the benchmark, along with the attribution of this excess return to our tactical allocation decisions and our manager selection decisions. This is the only way to substantiate our long term claims and to validate that we have delivered on our investment process.

    As to annualizing attribution for less than a year: two thumbs down on this. Just as it is unethical to create performance out of thin air by stretching out a short period to a year by annualizing, it is equally prohibited to manufacture excess returns that haven't been earned.

  2. Thanks for your comments; this will help with my analysis.

  3. From Andre Mirabelli:

    If you recall, I gave a presentation at one of your conferences on this very topic and at the time you said that you were going to mull over my explicit approach to the annualization of all properties, including attribution values. I also contributed a chapter on the exact same same topic to one of Carl Bacon’s books.
    If attribution results (say, for different funds) are to ever be compared for periods of different durations, either for longer or for shorter than a year, then annualization is necessary.
    As to the issue of smoothing out information, I find that this is best solved by presenting, as is done by Opturo, a graph of the cumulative impact of each attribute over time. This does not require annualization and also provides a visual measure of the volatility of the attribute values. However, while such graphs are very informative for a single fund, they do not provide a ready basis for comparing funds.

  4. Andre, thanks for sharing your thoughts. I vaguely recall our earlier discussion, but not the details; sorry. I will take into consideration your points as I ponder this further; thanks!


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