Wednesday, March 20, 2013

Mutual funds, ETFs, and Attribution

One of our GIPS(R) verification clients asked me recently about attribution and mutual funds/ETFs; whether there were special requirements for them.

Some thoughts:

Remember that with attribution, we align "sets" of securities in the  portfolio with their equivalent "set" in the benchmark. An aside: to my knowledge, I may be the first person to use the term "set." I think it works, though. For example, we can have a:
  1. Set of all items: this would be the portfolio
  2. Set of one item: this would be security level attribution, which I oppose, save for absolute attribution (aka "contribution")
  3. Set of no items; i.e., the "null set": would apply where the portfolio is void a set (e.g., a sector) that the benchmark has securities in.
We group securities into sets (where sets can be asset classes, sectors, subsectors, industries, etc.), and then match what's the the portfolio (that is, the weights and returns of each of the sets) with the corresponding set in the benchmark.

Thus, if we have mutual funds that fit into the small cap equity set, for example, then if we can group the securities in the benchmark by market cap, they would align. Nothing really mysterious here.

A second important point: unless the manager is MANAGING the mutual fund, we do not drill down to the security level, when it comes to evaluating the manager's performance, since he/she did not pick the underlying securities.



  1. Some plan sponsors and consultants do drill down into the fund/commingled fund to ofter a more realistic exposure vs treating the holding as one asset. The abinito processing makes such analysis possible and was infact being done at Bankers Trust back in 1990's. The fund/commingled fund owns the shares but the investor is a owner of the fund although sometimes funds/commingled funds can provide actual shares vs redeeming and proving cash and incurring brokerage and sec fees.

  2. Mark, thanks for your comment. This raises a couple points. First, you're looking at it from an "exposure" perspective, not attribution. To do attribution on the underlying sectors or securities does nothing, in my view, for the investor who purchased the fund; rather, it reports on the fund manager, which is fine, but is a different set of questions. Second, the information from publicly traded mutual funds is normally stale; funds are, I believe, prohibited from providing it in a more timely manner to a limited number of investors, as all would be entitiled to it. Nevertheless, this can be interesting information, I'm sure.

  3. Thanks for bringing this up Dave. I think attribution becomes confusing if you are an asset allocation manager and you have a mix of geographies, cap sizes, and sectors in your portfolio. Take this sample allocation:

    50% US Large Cap Mutual Funds
    15% US Small Cap Mutual Funds
    10% Europe/EAFE Mutual Funds
    10% Asia Mutual Funds
    10% Global Small Cap Mutual Fund
    5% US Energy ETF

    Where do you draw your "set"? If you say geography where do you put a global fund? If you say cap size where do you put an all cap ETF? It seem like you would have to drill down to the security level.

  4. Chris, thanks for chiming in. Sorry for the delay in posting.

    I'm fine drilling down to the security level to CREATE sets, but I wouldn't evaluate the individual securities. ETFs is an entirely different matter, as are mutual funds. As the investor, you're picking the ETFs and the funds, not the individual securities, and therefore it would, in my view, be inappropriate to do the analysis at the lower levels, unless you had interest in evaluating the decisions of those who constructed and managed these pooled vehicles. Great topic that I may take up in another post ... thanks!


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