Edition: September / November 2015
Survey-data – a blunt tool
in the toolkit
When trustees need to select an investment manager, a first stop might be to turn to a readily-available manager surveys. But is this really the right starting point?
For portfolio construction and manager selection purposes, managers can be classified into two broad groups: Beta farmers and Alpha hunters. Beta farmers are set up to capture the characteristic benefits of an asset class as efficiently as possible. They tend to benefit from economies of scale and their performance is normally explained by the return of the asset class.
Alpha hunters are a more mysterious breed – the returns they achieve can't be fully explained as being derived from known factors i.e value, growth, and momentum. In this sense, alpha is skill-factor.
A problem is that many managers' returns are a function of both beta - and alpha-drivers.
A manager might be seen to be scoring well in a traditional survey for months on end, but does this make him, or her, an Alpha hunter? Or if a manager seems to closely track the benchmark, can this manager be used as a Beta farmer in constructing your portfolio? Surveys don't separate this neatly, nor do they generally claim to. Survey data is a raw input that needs to be synthesized to information that can be used for an investment decision.
Choosing a manager that has performed well over the last few months (or years) from a survey might result in picking a favourable theme and not necessarily a manager with skill. In this sense surveys give an indication of what investment strategies are in or out of favour. This information is not easily readable to a person unfamiliar with the managers' philosophy and process or who does not have at least a rough idea of the portfolio holdings of the fund. This information is indirectly in the surveys, but requires careful reading between the lines.
Extracting maximum information from a survey entails inferring knowledge onto it – knowledge that is gained from experience and extrapolation from other data sources. As such, it is easier to start a manager search by looking at the entire universe of potential funds, rather than just focusing on the survey managers. Pure performance measures (data most surveys are based on) would be a component of such a search. The focus would not just be based on just the simple return or on some basic single point-in-time ratio such as the Sharpe ratio. Instead, the managers' performance should be assessed relative to the real risk taken to achieve that return, with risk in this sense being more broadly defined than just standard deviation. The manager's might be outperforming peers or a benchmark, but taking tremendous risk. A survey would not convey this information.
An understanding of the attribution of the manager's returns would be essential. Was the manager over-weight in momentum stocks, small cap companies etc? How does the manager's different returns-per-factor compare to a generic factor? For instance what if the manager out-performed his peer group due to being overweight momentum stocks, but in isolation his momentum 'portion' underperformed the market (beta) in momentum-stocks? He actually destroyed value by not being rewarded adequately for the risk he took.
A survey could be used as a (non)sense check to compare a shortlist of managers against a peer group; as an outside quality control; and to ensure personal bias has not had a strong influence on the outcome.
Surveys are also useful to become aware of managers that might not have featured in the original universe you considered forselecting a fund. They remain, however, a fairly blunt tool in the manager selection toolkit.
For more information contact Petri Greeff on firstname.lastname@example.org
or call +27 (0) 21 673 6999