When it comes to analysing manager performance, quantitative techniques give mixed results. A qualitative approach adds more value to the manager selection process, writes Guy Ester at Insinger de Beaufort
The construction of multi-manager portfolios is one of the areas within asset management that is relatively new and as such is still in more of a formative phase regarding the research process than, say, traditional equity portfolios. Nonetheless, the manager selection process itself, obviously a key component of constructing a multi-manager portfolio, has already received a fair amount of attention.
The main driver behind this has been the role of the investment consultants who for many years have been employed by pension funds in both the US and UK, and to a lesser extent in continental Europe. The multi-manager therefore has some established ground within which to root his work, but at the same time ample opportunity to add value with further research. It is the aim of this article to provide insights into a number of important aspects of the multi-manager process.
An obvious starting point in a discussion about running multi-manager portfolios is the ‘generic’ manager selection cycle, as shown in Figure 1. If we break into this cycle at the inception point of a portfolio then we begin with the quantitative analysis step.
Quantitative analysis is one of the topics which has received the most attention on the manager selection side. The simple reason for this is that the possibility of identifying managers able consistently to provide above-market levels of return on a risk- adjusted basis is in direct conflict with the efficient market hypothesis, as proposed in 1970 by Fama. In the 30 years that have since passed, the efficiency or inefficiency of markets has still not been unequivocally established, thus providing ample time for the academic community to produce a large body of research on the topic.
The bulk of the historic information available on fund managers typically takes the form of time series of returns and therefore unsurprisingly nearly all the work has been quantitative in nature. Qualitative information is available to a degree, but is mostly centred around fund manager educational background.
As already alluded to above, the research has not come up with consistent conclusions. The early and seminal papers by Treynor1, Sharpe2 and Jensen3 published in the mid 1960s raised questionmarks behind the rationale of investors in mutual funds, since all three concluded that there was no systematic outperformance of benchmark portfolios.
Some of the more recent literature, on the other hand, contains reports of both risk-adjusted outperformance of mutual funds and, significantly for those in the manager selection business, persistence of those returns going forward. Unfortunately, the evidence is not conclusive and there is therefore still no general acceptance of these facts. It is not the intent of this article to go more deeply into a summary of these studies, however, interested readers will find the articles of Carhart4 and Malkiel5 useful starting points for further reading.
From our own perspective, which has been formed both by studying the academic literature and proprietary research, we would note the following two points:
q A majority of the papers published agree on at least one thing – quantitative analysis may be used to weed out persistent losers. Therefore at the most basic level quantitative analysis may be used as a coarse screening mechanism.
q Even though some studies report quantitative methods of selecting future superior performing funds, we prefer not to rely wholly on mechanical systems, which may not offer the same performance going forward.
Rather than focusing purely on the conclusions of the literature, often the most value may be added by inspection of the methodologies employed to analyse the managers. Obvious examples are the three commonly used performance measures in the papers already mentioned above (Treynor ratio, Sharpe ratio and Jensen’s alpha) all of which deserve a place in any quantitative manager analysis. More recently the work of Sortino (downside risk) comes to mind.
Another important element in time series analyses of managers is to recognise that statistics need not remain constant over time. For example, rather than focusing purely on whether a manager is high or low beta, it is more interesting to follow the development of the fund’s beta over economic cycles. A good active manager, for example, might choose to increase the beta of a fund during a favourable economic period, and lower it during a more negative climate. Another example is the concentration of relative returns over time. Usually it is preferable to see a gradual accumulation of outperformance rather than in several discrete ‘bets’.
Our general conclusion on quantitative research is that it is highly useful in characterising a fund, and in providing a solid basis for discussion with the fund manager. However, utilised alone it provides a dangerous foundation for predicting future outperformance.
On the basis of the above we have therefore developed a manager selection process which is based around the tenet that the most value to be added in manager selection is not quantitative, but qualitative in nature. In terms of the manager selection cycle shown in Figure 1 the qualitative analysis refers to the manager analysis, but also more generally to his/her investment environment.
Of the many topics which constitute a comprehensive manager selection programme in this article we touch upon two important areas: teams and investment barriers. As a precursor to this discussion we should also mention that the main forum for assessing not only the topics surveyed in this article, but more generally, is a personal discussion with the manager. Whilst most asset management houses provide useful background literature covering the investment teams and processes, in our experience many of the more detailed workings of the process will not be covered adequately. In addition, a single meeting is often not sufficient to establish the type of relationship which goes deeper than merely scratching the surface. For this reason we would emphasise that close contacts with the managers are of primary importance in running multi-manager portfolios.
Within the asset management discipline a diverse range of team sizes, structures and dynamics are present. Often these variables are deeply embedded in the whole culture and investment approach of the asset management house in question, and as such deserve an important role in any serious manager selection program.
In the field of study known as ‘group dynamics’ four phases of team interaction are commonly distinguished. These comprise:
q team formation;
q strategy development;
q information processing, and
q project documentation.
Within the first phase, that of team formation, the raison d’être of the group is established, and the individual relationships and influences are founded. An essential component of this phase, which may or may not lie in the team’s field of influence is the size of the team. It is this aspect which will be discussed here.
An important input in determining the optimum team size is the nature of the task to be accomplished. The team structure of 12 which was utilised so successfully in the Japanese production industries is not necessarily indicative of the ideal team size for running a mutual fund. Broadly speaking there are certain task attributes which suggest a larger team and others which favour a smaller team.
If the nature of the task is primarily oriented towards manual labour or is dependent on information gathering then a larger team size is beneficial. Such tasks may also frequently be characterised as relatively simple or routine and often require lower levels of interaction with other team members.
If the task is more complex or abstract then often a smaller team is found to be beneficial. This is due in part to the requirement of complete communication between team members. The impact that team size has on this variable is illustrated in Figure 2.
In Figure 2 the relationship between the team size and the communication channels is illustrated. Note that between any two team members there exists not one but two communication channels, underlining the fact that communication consists of two way traffic. The point to note is that even for relatively small team sizes such as 10, almost 100 communication channels exist. This gives rise to one of two options – either considerable resources need to be committed to communication channel maintenance or a significant amount of information leakage must be tolerated.
In relating the above to asset management we restrict ourselves to two broadly typical situations for the sake of brevity. In the first case a team makes all the decisions more or less collectively and draws on other sources for informational purposes. In the second case the decisions are delegated to sub-teams (for example a currency team, a sector team, asset allocation team etc) and then brought together to form a portfolio.
Starting with the case of a single team which makes all decisions. Here communication is paramount and therefore from this point of view the team should be as small as possible. Additionally Katzenback and Smith6 make the point that “small size (less than 10) is admittedly more of a pragmatic guide than an absolute necessity for success” but add that one of the major problems with large groups is that they “... are more likely to break into subteams rather than function as a single unit”. We feel this to be even more so within the context of top-level asset managers due to their above average levels of intellectual development, and corresponding confidence in their own outlook, characteristics which do not naturally lend themselves to team participation. A final point to note is that as team size grows individual member dissatisfaction tends to increase again pointing to a small team as preferable.
In the case of reducing the decision making process to several sub-teams there are number of pitfalls which may reduce effectiveness. Firstly, within each sub-team the same constraints as above apply – the number should be as small as possible to achieve the task (ideally no more than five). Secondly, the very fact that sub-teams are utilised is an acknowledgement of the fact that the entire group cannot communicate effectively in its entirety. Therefore information leakage has been accepted, and the structure is such that the effect of this is minimal regarding the effect on decision quality. In practical terms, such a structure would for example be suitable in the case of a team of sector specialists each running their own sector portfolio and then combining these to form a single portfolio. A less workable structure would be one where teams are formed by currency, asset allocation, macro themes etc. In the former case the information leakage may have some impact on performance, since perfect communication between sector specialists might have some influence on the final decisions when compared with little communication. In the latter case, however, the information leakage may well prove critical since the team topics are so interdependent. In this case therefore forums would need to be set up to increase the number of communication channels.
We use the concept of investment barriers when studying any managers’ investment process. With the term barriers we refer to any artificial barrier which the managers impose on themselves and thereby restrict their investment opportunity set. A few examples will highlight this problem. One of the most well publicised and broad restrictions which asset management houses impose on themselves is that of style restriction. A house that proclaims itself to be a value investor, and is therefore also selected on this basis, frequently cannot invest in growth type stocks even if it feels that these are superior in the current climate.
The problem is exacerbated by investment advisers/consultants who screen portfolios on the basis of simple valuation ratios to test a portfolio for style bias. If a manager is found to deviate from these simplistic criteria he or she may well be fired even if there is no performance-related basis for doing so. Therefore deviating from style becomes too much of a business risk and the house imposes restrictions on investments at the management level. What we judge to be important in an investment house is a certain degree of flexibility or pragmatism. Obviously one does not want to hire a value style manager and base a portfolio construction programme around certain assumed performance characteristics, only to find that the manager has created a pure growth portfolio. However, at the end of the day it is superior returns that everyone is seeking from their investment manager. In fact the whole process of manager selection is based on the fact that superior managers can be identified, and should therefore exhibit superior stock selection capabilities within the context of the prevailing economic conditions. Note that again we are not referring to managers who take large timing bets, but rather make gradual shifts away from strict valuation criteria. It would appear rather self defeating to engage in complex manager analysis in search of the best, only to eventually demand a standardised, as opposed to superior performance.
Investment barriers such as style restriction are obviously well known, however the same idea may be used at more detailed levels of the investment process. For example many managers utilise some kind of automated screening process, econometric models or even human filters to produce a shorter list of stocks. Again what we judge to be important is the level of balance the manager is able to achieve between running an organised, well monitored process – which is a positive attribute – set against the degree of dogma and idea stagnation induced. Creativity is an essential characteristic of a top asset manager, however it cannot be allowed to run wild. Setting up the correct process to contain the creative element is a difficult task which is faced by all asset managers. It is the role of the manager selector to determine the degree of success achieved.
In conclusion, it is worth reiterating that the whilst the multi-manager business already has a number of accepted paradigms, none of these are ‘set in stone’. Given the space available we have given our stance on several important topics, but obviously many more exist. The challenge for the multi-manager industry going forward is to continue to refine and redefine the existing models such that the degree of confidence we have in selecting the winners of the future is maximised.
Guy Ester is head of research at Insinger de Beaufort in Amsterdam
1 J Treynor, Harvard Business Review, 43 (1965), pp 63-75
2 W. Sharpe, Journal of Business, 39 (1966), pp 119-138
3 M.C. Jensen, Journal of Finance, 23, (1968), pp 389-416
4 M.M. Carhart, Journal of Finance, 52, (1997), pp 57-82
5 B.G. Malkiel, Journal of Finance, 50, (1995), pp 549-572
6 J. Katzenback & D. Smith, Harvard Business Review, March/April (1993), p 114