One suggestion is that benchmarks should be:
- transparent & unambiguous
- frame-able & customize-able
- appropriate with full coverage
The source of this suggestion is Setting the Benchmark: Spotlight on Private Equity. This was discussed by All About Alpha. The paper considers indices and peer groups as benchmarks. They did not consider random portfolios.
Let’s look at their criteria in the light of random portfolios as a possibility.
There are two ways that random portfolios can be used for benchmarking. The fund manager can generate the random portfolios themselves. The advantage of this is that they will have the relevant data to hand, and they know their universe and their constraints.
Alternatively, investors can generate the random portfolios. The advantage of this is that they don’t have to worry about the fund manager having their thumb on the scale. The disadvantage is that they will need to guess about the constraints and possibly the universe.
Transparent & unambiguous
The paper wants the transparency to pertain to:
- components (assets)
All of these will obviously be true for investor-created random portfolios. They may be only partially true for manager-generated random portfolios — but generally at least as transparent as other methods of benchmarking.
Frame-ability & customize-ability
Frameability or measurement is the ability to clearly understand what the interpretation of the comparison means
Using this definition I claim that random portfolios are kilometers ahead in frameability compared to an index or a peer group. Comparison against an index is very noisy and will generally take decades to get a statistically significant result.
We never know what a ranking in a peer group means. If all the peers have similar skill, then it is a ranking of luck. Nothing to do with skill at all.
Customize-able has to do with creating sub-benchmarks. This is, of course, possible to do with random portfolios.
The shadowing method of using random portfolios is a particular form of customization. We customize to the portfolio that the manager holds at the beginning of the time period. This is the real problem that the manager faces. The portfolio at the start of the period is a given. Shadowing provides much more precise answers than other methods.
Appropriateness & coverage
Is the benchmark representative of what the manager is doing? Again random portfolios shine for this criterion — they are representative of the constraints on the manager as well as the asset universe.
The final criterion is investability. I’m yet to be convinced of why an investor should care.
But if an investor does care, then the possibility that jumps out of the random portfolio world is to select the minimum volatility portfolio among those that meet the constraints.
The “Setting the Benchmark” paper is particularly interested in illiquid assets. Random portfolios are not especially appropriate when prices are not readily available. However, it is hard to imagine that anything will produce much information in such situations.
Is there a different set of criteria for benchmarks that is better?