Friday, January 24, 2014

10 Key points about rates of return

At the core of performance measurement is the rate of return. While we are often distracted by attribution, risk, and the Global Investment Performance Standards (GIPS(R)), the ROR is fundamental. And so, there are some key points we should agree on.

1. Use time-weighting to judge the manager only when the manager isn't controlling the cash flows. That's it! No other need for it. While it seems to be the virtual universal way, that's sadly just how it's worked out, but fortunately a lot of correcting is going on, and many are rethinking the universal application of a method that was only intended to serve a single purpose.

2.Use money-weighted returns to judge the manager when the  manager controls the cash flows. Regrettably, GIPS' narrow application of the requirement for money-weighting only does part of the job. The rules are more complex than they need to be, and are often confusing. The simplest rule: if the manager controls the flows, use money-weighting. Hopefully, that'll be adopted worldwide (even in GIPS!) some day.

3. Use money-weighting to show how the client did. Whether the client controls the flows or not, money-weighted returns are the way to show the return of the client. More and more are realizing this, including the GIPS Executive Committee (with the introduction of the guidance statement for plan sponsors (i.e., asset owners)), GASB (the U.S. Government Accounting Standards Board), Canadian regulators, and others. 
Let's keep it going!!!

4. Know when to aggregate and when not to. If you have several accounts and want to know how your managers (collectively) are doing, in theory you have two choices:
  • Provide a time-weighted return on the aggregation of assets
  • Asset-weight the individual account time-weighted returns.
I vote for the latter. Why? Because the aggregate isn't being managed by a portfolio manager. It's essentially the same argument I have against the aggregate method for composites: no one is MANAGING the composite! The return is meaningless.

AND, if these managers include ones that control the flows, I still vote that you asset-weight, but for those managers that have money-weighted returns, use the MW returns, don't switch to time-weighted. The result is how your managers are managing your portfolios ... this can have some value. The return of the aggregation is valueless (that's in addition to being meaningless (see above)).

If you have several accounts and want to know how you're doing (collectively), you have two options:
  • Provide a money-weighted return on the aggregation of assets
  • Asset-weight the individual account money-weighted returns.
I vote for the former. I think it will be a better representation, since it takes everything into consideration.

5. Shorts are special, they need their own linking method. If a portfolio holds short positions, the standard geometric linking won't work; there's a special adjustment to that approach which I've highlighted in the past, which should be used.

6. Speaking of shorts, don't net your longs and shorts. If a portfolio holds both long and short positions of the same stock, or periodically goes back-and-forth between long and short, keep them separated. While it IS possible to derive a return across the time period, it's better to segregate them so that you can show the return of each side, as they're intended to do completely different things and reflect differing views of the market.

7. Show returns to two decimal places. Two decimal places mean you're showing to the basis point (1/100th of a percentage point). To show to only one decimal place doesn't provide enough precision; to show to three is overly detailed and unwarranted. Are there exceptions? Yes, but they're rare.

8. Select the right time-weighted return formula. Ideally, use the exact method (aka "true daily" and "pure"). If you have to use an approximation method (e.g., Modified Dietz), revalue for large flows, and set your "large" threshold to no more than 10 percent. In addition, if you go with this latter approach, test flows against the  most recent revaluation point.

9. Select the right money-weighted return formula. We've helped firms implement money-weighted methods, and have done  so using Modified Dietz, as an "approximation" to the true, money-weighted return (aka, the internal rate of return (IRR)), but recognize that this is less than ideal. The preferred approach should be to use the IRR. While the Modified Dietz is often quite close, as with its ability to approximate the exact TWRR, it sometimes lacks the accuracy we'd like.

10. If you're not comfortable explaining why a client has a positive TWRR but lost money, figure it out. No, it's NOT wrong; no, it DOES make sense. You need to understand WHY the return is this way and be able to explain it in an intelligible way. Step one is to identify the cash flows (there has to be at least one) and step two is to look at the performance pre- and post-the flow(s). That's probably all you need to do; if you need help, contact me.


  1. Stephen Campisi, Modified Dietz afficianadoJanuary 24, 2014 at 8:37 PM

    A very useful compilation of reasonable guidelines in the appropriate use of rates of return as a measure of... what? Is a return really a measure of success? Maybe for a fund manager if investing was really a "horse race" - but even then, there's no mention of risk. This isn't a criticism, just a comment that rates of return are only one part of the puzzle. Back to kudos to you: if we are to assemble the mosaic of performance and turn it into an effective story of what happened, we MUST select and apply the variety of returns appropriately, and your guidance helps greatly.

    That said, you may want to tweak (and not twerk) the 8th and 9th "commandments" you've put forth. Logically, something can't be both itself and the antithesis of itself, and so Modified Dietz can't be both time-weighted AND money-weighted. You may buy into the idea that a time-weighted return can be created by linking money-weighted returns, but this is a specious argument whose only defense is that it's "a law" that was handed down through the GIPS standards. That doesn't give it any intellectual credibility.

    Finally, there must be a purpose for each return and you make this point clearly and effectively. With that in mind, you should get off your "high horse" of making a "distinction without a difference" by contrasting Modified Dietz with the "true" internal rate of return. In real life there is rarely a material difference between these results, and you fail to mention the strongest value of the Modified Dietz version of the IRR: it provides BOTH a return and an internally-consistent weighting that allow the complete and accurate explanation of the change in wealth for the asset owner (the "forgotten party" in the investment game.) It's simple, intuitive, it provides only one answer, and it is accompanied by the required weighting that's needed to put it to good use in explaining performance. This is like gold in the return game; you would do well to "give it its props" and stop treating it like the "crazy Uncle" or "redheaded stepchild" of performance.

  2. Oddly, Modified Dietz CAN be both time- and money-weighted: it depends on the implementation. In both cases it's approximating the TRUE measure. For time-weighting, we link subperiods; for money-weighting, we derive the return across the period with no linking. While Mod D is often a very good proxy for IRR, there are times when the results can be QUITE wrong (I've highlighted these in the past, I think).

    This wasn't intended to be EVERYTHING (thus, no reference to risk), but rather to touch on merely one piece of the rather sizable puzzle.


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