Security level attribution: whether you need it or not

I suspect that software vendors are often frustrated by some of the requests they receive from clients. Clients occasionally ask for things that don't make sense, but the vendors have to respond (or attempt to convince the client that they're wrong, which might be a challenge ("a man convinced against his will, is of the same opinion still")). Take security level performance attribution, for example. Attribution software RFPs will typically ask the vendor whether or not they provide this. If they answer "no," they may be excluded from the search. But why does anyone need this? I have argued against security level attribution for some time. As with many aspects of performance and attribution, there's controversy surrounding this topic, so I realize that some will argue that security level attribution makes sense; I just don't see that it does.

There are three scenarios when it comes to securities:

1. The portfolio owns the security but the benchmark doesn't
2. The benchmark owns it but the portfolio doesn't
3. Both the portfolio and benchmark own it.

I've shown previously (in articles and our newsletter) that the "Brinson models" are somewhat flawed when it comes to dealing with cases when a security is in one (the benchmark or portfolio) but not the other. Granted, there are "workarounds" available to deal with this, but the models themselves don't come equipped to handle these scenarios. But if the portfolio is invested in a security that's not in the benchmark, is this an allocation or selection decision? Surely it's selection, yes? And if we're both in the same security and I own more, is this an allocation decision? HARDLY!

Let's consider the storied S&P500(R). It, as the name suggests, is comprised of 500 securities, allocated across 10 GICS(R) (Global Industry Classification Standard) sectors. The largest sector today is Information Technology, which comprises roughly 19% of the index and has 76 securities. Let's say that I am managing against the S&P500 and for technology decide to underweight. And, for this sector I select just three securities: Apple, Microsoft, and IBM (all members of the S&P500). At the sector level, my portfolio will be evaluated on allocation (I underweighted) and selection (my composition differs from the index's). At the security level, I would no doubt have more in my three securities than they have in the index, given that these three must share the allocation with 73 other securities. Is my overweighting going to be evaluated? For what purpose? I underweighted the sector; I'm not overweighting the securities! It's all about selection!

At the security level, contribution or absolute attribution should be employed: not relative attribution!