The selection of the benchmark that will be used throughout the investment process is an important is...
Benchmarks are used in a wide range of ways. Initially, it is used to understand the risk and return properties of a particular asset class.
Once the asset allocation has been decided, significantly based upon this analysis, the manager selection decision is made.
This decision is typically based upon the ability of the selected manager to add value compared to the risk/return profile of the asset class, again using the benchmark to proxy the asset class concerned.
Finally, the success or failure of the asset allocation and manager selection decisions are assessed, again using the benchmark as the proxy for the asset class.
Throughout this process, the assumption is made that the benchmark accurately represents the asset class as a whole. All benchmarks that are broadly diversified are assumed to represent the asset class and, essentially, to be interchangeable. Is this the case?
The answer to this question is not always yes. Most broadly diversified portfolios will, indeed, produce similar returns at the total fund level. The volatility of these portfolios will also be similar.
However, when managers' added-value can often be measured in basis points, and when selection decisions can turn on basis points in fee differentials, differences between alternative benchmarks of basis points of risk or return do matter.
Such differences certainly exist and there are two basic reasons for their appearance. The first relates to the fact that benchmarks have traditionally been constructed using sampling methodologies.
Originally this was required because of the logistical problems related to measuring 9,000 companies every day. With the technology available 30 years ago, this represented a significant challenge.
Today, however, although this process is time consuming, it is now achievable.
Sampling methodologies always introduce errors, and when they are not required they should not be used, as the errors (and therefore the risk) that they introduce are not compensated for by the market.
Full replication allows for exact matching of the risk and return opportunities available in the asset class as a whole.
The second reason is that standard methodologies use the total market capitalisation of the companies in the index, rather than the proportion of the equity actually available for purchase.
Again, this was originally a simplifying assumption. However, it introduces error, which is not compensated by the marketplace. With around 30% of the world's equity capitalisation unavailable for purchase, this is neither a minor effect, nor an academic one.
The effect on the asset allocation policy of the plan of these errors can be quite significant. With around 50% of Japan's equity market not available for purchase, adopting a simple total market-capitalisation weighting methodology is effectively making an extremely substantial active bet in favour of the Japanese equity market.
Sometimes this will help the fund's performance. At other times it will damage it. It is, however, an active bet, and it affects every stage of the process, from allocation, through manager appointment, to continuous manager assessment.
The nature of these active bets has to be understood. Active bets are also caused by the selection process, with the behaviour of the index affected only by the stocks included and not by those excluded, no matter how well or poorly they perform.
In addition, the regular rebalancing caused by the active selection process is costly and the fund will incur transaction costs on every occasion.
Each of these factors introduces error into the process and causes the benchmark to misrepresent the asset class. Each misrepresentation is minor, in percentage terms, although they are certainly of the order of magnitude of manager fees.
The most important thing is to understand the active bets being taken during the benchmark selection process. To do this requires careful analysis of the options, the risks and the costs involved. Not all indices are the same.
Ian Toner is a vice-president of the Equity Index Group at SalomonSmithBarney
Slow progress in improving diversity
Share purchase deal with assets of £28m
Came into effect in January
Three examples of compensation rule issues
Buying in baskets