My response was that they should ideally rebalance the benchmark when they rebalance the portfolio; to rebalance more frequently would detract from the variations which are being caused within the benchmark which would most likely vary from what they're doing, and thus eliminate the allocation shifts which may or may not be in the manager's favor. But to confirm my beliefs I reached out to Neil Riddles, aka "the benchmark king" to get his thoughts. He responded:
One of the important considerations for benchmarks is that they should be a passive representation of what the manager is doing in the active portfolio.
If the portfolio has a set rebalance period, then mimicking that makes sense. For example, if it is rebalanced quarterly, then rebalancing the benchmark quarterly is best.
If the portfolio is rebalanced when it gets a certain percentage out of line, that policy should be applied to the benchmark as well. For example, if the portfolio is rebalanced when it gets out of line by 10% (55/45) then the same rule ought to be applied to the benchmark.
Ideally, the benchmark rebalancing should not be dependent on what goes on in the portfolio. So, if the portfolio rebalances on a non-regular schedule (or does not follow other explicit criteria) then we would not want to rebalance at the same time. The reason is that, absent an investment in the portfolio, the investor could not achieve the benchmark results.
At any rate, rebalancing daily is likely to maximize the error introduced because the benchmark is trading costlessly.
I think our views are almost identical. Formal guidance would always be helpful, and we'll look to develop that at some point.
p.s. I dubbed Neil "the benchmark king" quite some time ago, using the nicknames assigned to various prisoners in "The Great Escape." For example, Steve McQueen was "the cooler king." Oh, and if you've never seen the movie, you should!
.png)
This is wonderful advice which should become a more important and visible part of every performance analyst's training. I recall that benchmark rebalancing is a key topic that was frequently discussed by Damien Laker; this was in my opinion one of his more important contributions to the body of knowledge of investment performance.
ReplyDeleteIn the query you cite we have yet another example of the wrongheaded notion that "more frequent is always better" - whether this applies to the frequency of return calculation, or attribution calculation or performance reporting - with benchmark rebalancing now jumping on this bandwagon that is heading in the wrong direction and picking up speed.
Here again we have a failure to understand the difference between "precision" and "accuracy." Answering the wrong question with precision is the opposite of accuracy. In this case, a benchmark that is rebalanced at a different frequency than the portfolio is simply inaccurate, because it is not representative of the portfolio to which it is compared. Just as you would not compare the portfolio to a benchmark with a different long term asset allocation, you should not compare a quarterly-rebalanced portfolio to a monthly or daily rebalanced portfolio. As I had heard Damien mention in the past, the definition of the benchmark should include the frequency of its rebalancing. Wise advice.
We all know (or should know) that rebalancing in a trending market hurts performance - you cut back on your winners and you keep reloading the losers. If you rebalance the benchmark more frequently in a trending market then you are "gaming" your excess return, since you are generating an artificially low benchmark return.
I faced this same situation with a consulting firm that calculates performance for one of the portfolios I manage - they rebalance the benchmark monthly, thinking this to be more precise. I argued that a quarterly rebalancing would be more accurate because that would represent our true rebalancing frequency for the portfolio. Who won? I did, since I'm the portfolio manager. The results? Consider the 1-year return for the period ending March 31, 2010. It was a very strong year for equities compared with bonds, with the examples of the S&P 500 earning 49.8% and the Aggregate Bond index earning 7.7% and Cash earning only 0.2%. Our globally diversified 70/30 benchmark earned 39.57% if rebalanced monthly but 39.79% if rebalanced quarterly. This difference of 22 bps would have falsely inflated "excess return" as a result of rebalancing the benchmark too frequently.
Is more frequent always better? Clearly not. In this case, more frequent is worse: it's not representative of the investment process, and it creates spurious excess return. Moving to daily benchmark rebalancing is even worse since it wastes precious resources that could be used for more important things.
Steve, as always some excellent insights. Yes, this perception of "more frequent is better" is often seen as something that everyone of course would agree with, and challenging it can be difficult. In his most recent book (The End of Wall Street) Roger Lowenstein speaks of the exact precision some firms provide with their VaR numbers as if this level of detail has any validity; if anything it gives the recipients the false belief that of course the numbers must be true if someone is going to go to such precision in reporting. There is definitely an opportunity for improvement in our reporting.
ReplyDeleteIf I mimic benchmark completely, different re-balance strategies between benchmark and portfolio will introduce active risk, which does not make sense.
ReplyDeleteI think that's one reason we should apply same strategy to the benchmark and portfolio.