Thursday, April 7, 2011

Attribution: a matter of perspective

While I often speak of returns being a matter of perspective (that is, whether we're speaking about how the manager did or how the portfolio or client did) to determine whether time- or money-weighting should be used, the same holds true with attribution. I was recently approached by a portfolio manager who wants to report attribution from a selection, country, and sector perspective. Okay, so right there we have three different effects we may want to consider. But it gets even more interesting. Let's consider this first approach:

Here we see that we are able to reconcile to the excess return, using either the country or sector effects. And notice that the sector's weights are relative to the portfolio, meaning that we are evaluating how each sector (as well as country) contributes to the overall excess return. And so, we see how our allocation and selection decisions worked out at the country, sector, and overall levels.

Now, let's consider the following:

Here we have grouped the sector data together, so that we are only focusing on how each sector contributed to the overall return. You'll notice that the portfolio's effects are different than what we saw in the earlier example, but that's because our analysis is now from the perspective of the sectors and the way the manager invested relative to them.

In our third example you can see how the sector weights are now relative to their respective countries:

We are now answering the question, how did the manager's actions at the sector level contribute to each country's performance?

And so, there are lots of ways to "slice-and-dice" our numbers, to provide us with different perspectives as to what is going on. It's important to understand that this is possible and to decide which way(s) makes the most sense for you. You may want to look at attribution from multiple perspectives, which is great, but understand what the information is reporting to you.

By the way, I have ignored the currency effects to make this presentation easier; perhaps we'll add this factor at a later time.


  1. Dave, you raise an interesting point. I am very particular about it: There exists only one "correct" perspective in performance attribution - the one the decision makers had ex ante. Performance attribution is about attributing ex post results to ex ante decisions. If we take the wrong perspective ex post, we get results that are not meaningful: a country breakdown doesn't make sense for a portfolio that is managed according to sectors.
    Most of the time, the "one-dimensional" breakdown in your calculations is applicable. In most real-world portfolios, decisions are simutaneously accross varios dimensions (countries and sectors). There exist performance attribution "models" that can produce meaningful results for such portfolios. I talk about them in my Performance Attribution seminar.

  2. Andreas, thanks for your input. I have two thoughts on this: #1, I agree that there technically only one "correct" perspective, which aligns with how the portfolio is managed. However, it is also often helpful to look at the portfolio from DIFFERENT perspectives, to gain insights into what is going on. Flexible attribution systems, which provide the firm the ability to "slice and dice" their numbers, to look at things from different angles (perspectives) can be quite beneficial. A manager (and their clients) can learn a lot about what's going on, by looking at the portfolio in different ways.

  3. I fully agree with you that flexible attribution systems are needed. Too many currently available systems do lack flexibility. I think that flexibility is needed to custom-tailor an analysis to real-world investment processes. The number of "different" perspectives is unlimited; I do not value them as highly as you do. On the other hand, I do value highly differing "perspectives" when it comes to factor models.
    My answer is also motivated by the fact that I do not agree to the prevailing "relativist" attitude in performance & risk analysis: not all risk-adjusted performances are created equal, not all performance attribution breakdowns are meaningful. It does not follow that there exist "single best" solutions. What is needed is guidance on which methodologies make sense in which context. Critera to choose the "more appropriate" methodologies do exist.
    I am also very much opposing standardization: competition is what motivates people to come up with ever-better solutions. Standardiation typically results in reversion to the minimum, which is not in the interest of the client when it comes to performance and risk information.

  4. Andreas, thanks for the follow up.

    As for standardization, I DO support it, but not from a highly prescriptive perspective (that word again), but rather from a DISCLOSURE one; that is, to indicate HOW the firm does attribution which MODEL they use, whether it's TRANSACTION or HOLDINGS based, etc. I think there's value in this. Given the multitudinous ways and options available, some disclosures are in order.


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