What price complexity?
Over the years I have become a firm subscriber to the KISS school of investing. This is not some vague reference to a quantitative trading scheme developed by Paul Stanley and Gene Simmons et al, but rather the notion that investment is not a complicated game. The mantra of “Keep it Simple Stupid” applies probably as much to successful investing than any other paid endeavour.
Investing is not complicated. That’s not to say it isn’t hard, and doesn’t require skill, but it is not complicated. Unfortunately our industry – and those that surround it - earns a good deal from making it just so.
Prima facia examples of this include the whole “structured debt” class of securities (take stuff you can buy already, repackage it, charge a fee and sell it), hedge funds and especially hedge fund of funds (charging a lot for what in aggregate hasn’t looked any better than a conservative balanced fund), and private equity (charging you a multiple to leverage what you could already buy – and probably already owned).
Outside of these more obvious examples the complexity theme has worked its way into fundamentally quite simple structures. The whole “segregation and specialisation” of the traditional asset classes and the move away from simple balanced structures ultimately represents a move to greater complexity. While sometimes complexity can bring benefits, more often than not in our view it increases costs and can hide risks. There are often many stakeholders who can benefit from complexity, but not always the end client. Within Australian equities the last 6 years has witnessed a doubling in the number of fund managers (and a multi-fold increase in the number of products and strategies offered). At the same time we have seen more managers appointed to increasingly smaller and more specialised mandates (e.g. boutiques).
The net result has been higher fees paid to an industry where talent has become increasingly dispersed. We would argue that this is not in aggregate a positive result for end clients.
The purpose of this paper is to stimulate some greater debate within the industry by reviewing the actual reported experience of Australian equity managers and, on the basis of some realistic assumptions, set out some alternatives for the industry to consider in the construction of Australian equity multi-manager portfolios. While some of this analysis may be relevant to other parts of the portfolio (and again at the minimum we would hope stimulates some further analysis and debate) given the nature of the Australian equity market we would be a little careful in making broad assertions that these results can be more generally applied.
We would also be early to point out that as a large manager of “core” Australian equity portfolios some of this analysis may come across as self serving. We would make no apologies for that – the data is what the data is. In addition, after many years of hearing the “benefits” of more specialised structures we hope this goes someway to offering an alternative point of view.
There are a number of issues that should be addressed up front in analysing the data for reported performance of Australian equity managers. Firstly, there is no truly comprehensive source for data on manager performance in Australian equities. We have chosen to use the Mercer MPA Australian Shares universe but would be careful to point out in advance that this is not a full data set.
Secondly, in looking to analyse the performance of aggregated multi-manager portfolios we would also point out that a single comprehensive source of data on these portfolios doesn’t exist (note that in referring to multi-manager portfolios in this document we are considering any institutional combination of Australian equity managers with the aim of outperforming the index. This encompasses retail-multi-manager funds, implemented consultants, superannuation funds and other large institutional investors).
There are two very important points here that are worth making:
- While the main surveyors attempt to remove selection bias from their data, it is not possible to completely remove it. (You can’t force a manager to report data, and if they are on their way out they will simply stop reporting. Likewise, managers choose which products to add into the survey, and when they add them). For example, the Mercer Survey had 59 long only managers in December 1999. In December 2004 there were 45 managers with a 5 year track record and in December 2008 only 34 managers with a 10 year track record. While the demise of some of these managers is captured in the data it is not fully captured, and in any case a 50[(-60) was not found] survivorship rate implies some not insignificant costs in rotating away from “failed” managers.
- The reported data shows percentage value add, not what really counts which is dollars of value add. ie. A manager with $100mn receives the same performance weight as a manager with $5bn – yet the economic consequences of the two are quite different. In fact the point that the survey’s are not in aggregate asset weighted could be a part explanation for why the industry overall appears to outperform the index – something the passive side of the industry would argue is impossible over time (it’s a negative sum game after fees for all participants).
The simple model
Fundamentally, the simplest model for constructing an Australian equity portfolio would be to replicate the index - assuming, and this is in our view a big assumption, you believe the lowest “risk” portfolio is the index portfolio. Leaving that debate aside for another day, we will take the low risk/low cost option as being the index portfolio.
However, for those of the view that 1) active management can add value (after fees), or 2) charge an “active” fee to their end clients and therefore feel compelled to choose at least some active managers, consideration needs to be given to how to combine one or a number of managers to generate post fee alpha. We would set out that the “objective” of most multi-manager Australian equity combinations is:
“to deliver the highest post-fee alpha for a given level of relative (to the index) risk”.
In particular, the level of post-fee alpha targeted has to be at least sufficient to cover the multi-manager’s own internal costs or work involved in constructing the portfolio. Risk in this metric is determined to mean “variability” of the alpha, particularly negative variability.
Given the need (or desire) for a certain minimum level of alpha and the consequence that to deliver this some level of risk must be taken, the question can then be split into:
- What factors contribute to maximising post-fee (including tax and transactions costs) alpha; and
- What factors contribute to minimising negative variability.
Before conducting our own analysis it is worthwhile commenting on some of the available published reports on construction of multi-manager portfolios.
– Gallagher and Gardener (2005)1 make the point that in a study of US managers “We document increased difficulties for blended portfolios in their attempts to outperform the market. This arises due to significant erosion in the active bets of stocks held in blended portfolios. Our research also identifies a potentially significant efficiency problem for blended portfolios, where up to 15 percent of stock trading in any quarter is executed between common equity fund managers for a 10 fund portfolio case. As expected, this effect is amplified when an analysis is performed at the industry level. Improved efficiencies in active portfolio design would be most likely achieved where a single agent is permitted to construct and maintain the portfolio structure in a manner which does not erode the active investment opportunities offered by selected institutions.”
– Eggins (2005)2 reviews the Gallagher and Gardner (2005)3 Australian study and makes the point – that blending managers reduces risk, but with decreasing economies of scale. In addition the case is made that alpha does not decrease as the number of managers increase. This is shown in the chart below. We would agree with the first point (and as noted there are a number of other studies that point to the same effect, including Brands and Gallagher (2004)4). However in respect of the second point while the no loss in average alpha will be the case before fees, we would counter that given that most managers scale fees down with increasing assets the post-fee alpha will decline as more managers are appointed to smaller mandates. Eggins also makes the point to which we would agree that “Compared to a single manager, however, multi-managers do not incur any additional transactions costs.”
– Eggins and Parish (2007)5 make the point that there is a “positive relationship between excess returns and tracking error” based on the 5 year period to June 2006. They also state that “this result is not unique to this time period or asset class”. While that appears to be the case in some periods our results contradict this somewhat for later periods. In particular our results show that to the extent there is some pick-up in alpha with tracking error this is not sufficient to offset the higher fees.
The studies above and numerous others which measure the relative differences between boutiques and “institutions” or satellites and cores all broadly point to a number of outcomes with which we would intuitively agree – that beyond a point (say 5 or 6 managers) adding additional managers will not materially improve the risk return trade-off. However we would also point out that all of these studies focus on pre-fee analysis. Given the likely scale down in fees with asset size we would contend that the picture changes somewhat on a post-fee basis. We explore this further below.
The Risk/Return trade-off
One of the more common assumptions embedded in multi-manager portfolio construction is that when managers take risk, this risk will ultimately be rewarded (if not, your objective should be risk and fee minimisation). However, the question should be asked as to what extent this is correct. We have witnessed a strong rise in recent years in the demand for high-alpha (or should we say high “expected” alpha) mandates – which usually entail significantly greater concentration of stock ideas (and often higher turnover).
Part of the argument for such structures revolves around “capturing best ideas” and not “paying for an index” portfolio. While it sounds attractive in theory we would offer two thoughts for consideration:
- Superannuation funds (and most institutional pools) are long term investors. The greatest arbitrages in active asset management are typically time horizon arbitrages (taking a long term view will be rewarded). Higher turnover implies a demand for shorter-term returns. If you’re time horizon is 5+ years, this implies turnover of 20%. Many “high alpha” mandates have 100%+ turnover implying time horizons of less than 1 year. This is inconsistent with the notion of being a long term investor and trying to minimise total costs and taxes.
- In aggregate the superannuation industry in Australia (assuming an average 30% holding in Australian equities) represents over half the market capitalisation of the Australian share market.The notion that everyone can be in a concentrated, best ideas portfolio and have significant nonindex positions is mathematically impossible.
Ignoring fees, there are three basic possibilities that the risk/return trade-off could take:
– increasing returns to risk;
– stable returns to risk;
– decreasing returns to risk.
These payoffs are represented graphically below.
One could also postulate that you get a combination of these as you move along the risk axis (i.e. Initially increasing returns to risk, then stable, then declining). This would be our proposition, but firstly lets review the empirical evidence.
Analysing the actual pre-fee, pre tax results for active Australian equity managers from the Mercer MPA database reveals some interesting results.
The charts below show for three different snapshots in time the actual risk / reward trade-off for all enhanced index and long only managers in the Mercer MPA Australian shares universe.
Risk, in these examples, has been calculated as the “average over 3 years of the rolling 3 year annualised tracking errors” (or whatever longest period is available, subject to a minimum of 3 years). The attempt here is to get some reasonably realistic measure of relative actual risk through time. Alternative measures were also considered but didn’t lead to materially different results.
We have shown annualised excess returns over 3 and 5 year periods to 31 January 2000, 2005 and 2009. (Cleary our results in each case exclude managers with less than 3 or 5 year track records, which may in itself involve some (upward or downward) selection bias) – The appendix shows the individual manager results for each period over 3 and 5 years.
We can observe from the charts above that for the periods to 2005 there appeared to be a positive relationship between actual portfolio risk and return (especially in the 3 year data – albeit there are only 4 observations for that data point), this did not necessarily hold for the other periods. In fact, for the periods to 2009 returns initially improved to risk but then appeared to decrease as tracking errors rose above 4%.
This would to us appear to be a more intuitive result from a practitioners’ perspective. Given the relative narrowness of the Australian equity market (financials & resources making up close to 60%) while some increasing returns to risk are possible, as risk is increased substantially the degree of concentration in portfolios required (at either a sector level or a stock level) is such that portfolios become significantly more prone to individual stock or sector specifics, which are somewhat harder to forecast and therefore add to volatility but not necessarily to returns.
In addition, if we look at the performance of smaller cap stocks relative to large cap stocks (shown below) there appears to be some correlation with the relative performance of higher risk managers – both did better in the period to 2005 and correspondingly worse in the periods to 2000 and 2009. This is again not surprising as part of the increase in risk is likely to come through larger allocations to smaller cap stocks (one could argue “beta dressed up as alpha”). Albeit that the equal and opposite point could be made about high-risk manager performance in the other periods and therefore these are not valid points for comparison either, our rebuttal would be that if it is all down to small cap then why not just isolate this effect (and pay for it) separately. In any case, the reasons for the performance differentials could be the subject for future research.
Another way to consider this analysis is to look at the average reward per unit of risk taken – or the information ratio for managers over the same periods. This is shown in the charts below.
Again, with the exception of the period to 2005 we can observe the tail-off in reward to risk as risk increases above 4% actual tracking error.
So what does this mean from a portfolio construction perspective? The above results show that, in general, there is some reward to risk (in particular the average information ratio’s are positive) on a pre-fee basis, however this appears to tail-off as progressively more risk is taken in the portfolio. At the same time it would be our observation that fee’s for higher risk portfolios are greater than those for lower risk portfolios. We now consider the impact of fees on the overall portfolio.
Post fee analysis
In order to understand the likely post-fee impact of alternative manager configurations it is necessary to make some broad assumptions about likely fee levels for different mandate types.
For the purpose of this analysis we have made the following assumptions with respect to average fee levels. These fee levels are based on our own expectations and discussion with industry participants. Most critically, it is not so much the absolute levels that are important but the relativities between each type. We would agree that there is likely to be some dispersion around these levels from manager to manager, but for a combination of managers it comes ultimately comes back to the average fee for that group and the relativity between groups.
We have shown above two “expected alpha” scales. The first represents what would probably be the realistic expectation for managers of different types – and what could be argued represents a better than average actual outcome based on the value-add from manager research.
The second “actual alpha” expectation is based on the average alpha for different managers that has been historically produced as per the data presented in the charts above.
One area that makes this analysis potentially more difficult is the use of performance fees. At the single manager level it could be argued that the use of performance fees will potentially mean that higher risk structures can be put in place at limited overall cost to the client as if the performance is not delivered then it will not be paid for. This is something we plan on addressing more broadly in a forthcoming paper, however we would make the following comments:
– Properly structured, performance fees can help in achieving better post fee outcomes. However, this will typically involve lower percentage participation than is commonly the case, the use of high water marks (which is common in institutional mandates) and the need to retain managers for a significant length of time to allow the fee “benefit” from underperforming managers to offset the fee increase from managers who outperform. As underperforming managers are rotated out of the structure this potential future fee benefit is lost and aggregate fee’s rise.
– When an increasing number of managers are used (as is common in higher alpha structures for Australian equities) setting each manager with a performance fee generally results in aggregate fees rising – much the same as it would if you paid a single manager a performance fee for their correct individual stock calls but not their incorrect ones!
We have suggested in the fee table above a relative “flat fee” comparison for high-alpha managers. Clearly as fund sizes increase, given the fall off in average fees, different combinations of manager lineups will result in different aggregate fee levels. The chart below shows the total fee for different combinations of managers at different total fund sizes.
We can observe from the above that as mandate size grows, average fees (should) fall even in the more complicated structures. However, it is also worth pointing out that as size grows a “multi-core” combination is cheaper, or at worst broadly similar, to the passive + high alpha combinations in terms of total fees. This would change to the extent that:
- The passive component is reduced or replaced with enhanced passive;
- More “high-alpha” satellites are employed.
The charts below now take our assumed “alpha” expectations for each manager above and overlay this fee scale to generate a post fee alpha (although notably still pre-tax) for various combinations of passive, enhanced passive, core and passive + high alpha combinations plotted by changing fund size.
It is clear from the chart above that the post-fee outcomes now tilt significantly in favour of the option of using blended “core” managers rather than a passive+satellite (high alpha) approach. In fact, based on these fee differentials, the conditions required to make the multi-core approach similar to the passive+satellite approach are that the expected return of the core manager would have to be less than (by at least the fee differential) half of that of the high alpha manager. This is clearly not supported by empirical evidence.
While the results for total fees are probably not that earth-shattering (or shouldn’t be), what is of more interest to multi-manager portfolio constructors is what happens to total portfolio risk and post-fee returns as you employ alternative manager structuring arrangements.
Analysing risk and return
In order to analyse the risk and return of alternative combinations of managers we need to make a number of assumptions about the likely risk of those managers. We have assumed the following:
*It is clear that these numbers are better than historical averages, however we have (rightly or wrongly) allowed some benefit towards skill in manager selection.
In addition to this we have made the assumptions that:
– Alpha’s are normally distributed (albeit in practice they appear to have slight positive skew and higher kurtosis); and
– The alpha of any two passive managers is 100% correlated, any two enhanced passive managers is 50% correlated and any two core or high alpha managers is 0% correlated. In respect of this last correlation estimate we would note that 0% appears to us a low figure for any two managers consistently through time. We would content that while the actual data shows most managers alpha’s between -60% and 60% correlated6 given the limited breadth of the Australian market, as you increase the number of managers it will become increasingly difficult to find larger groups with low correlations. In any case, it is the difference between the core and high-alpha correlations that will matter not the absolute figures.
The charts below show the combined risk and reward of alternative manager combinations.
As we can see from the charts above there appears to be no clear risk and return benefit to selecting the “passive+satellite” approaches common in multi-manager portfolios today. In fact there are distinct benefits to selecting a smaller number of larger core managers. If we were to then consider the effect on tax from the higher turnover satellites and the increased governance required to monitor and research this larger group of higher risk managers, the benefits to selecting a smaller number of core managers appear to us to be relatively clear.
Australian equities is probably one of the “simpler” asset classes – certainly in terms of the access to data, managers and our familiarity with it. The concept of “keeping it simple” applies here as much as it does anywhere else.
However, despite the evidence, it is our observation that complexity and irrationality have crept their way into the structuring of Australian equity multi-manager portfolios much the same way they have in other asset classes and products. As outlined in this paper, we believe there are some simple approaches that could be taken to maximise the post-fee relative returns from this asset at minimal risk. We would note the following:
– We have observed a sharp rise in the demand for higher tracking error/higher fee mandates combined with lower fee/lower cost enhanced index and passive mandates;
– Such structures would make sense if the risk/reward curves were linear or increasing with risk, however the evidence shows that this is not the case;
– Our observations show that it is direct fees and indirect costs of turnover and taxes that increase with the level of risk taken, further reducing the benefits to “high alpha” structures;
– The information ratio has generally been maximised in the 2-4% “tracking error” range – a spacethat curiously has been reduced in recent years at the “expense” of these lower tracking error cores and higher alpha/fee mandates;
To this end, appointing fewer managers, to larger more diversified mandates with an after tax focus, at a point on the cost curve where average fees are minimised would result in better after tax and fee returns at lower aggregate risk. Such structures are simple, require less governance and lower monitoring costs and will make it easier for internal investment teams to aggregate and monitor total portfolio risk.
Complexity is over-rated.
The charts below show the results for each individual manager over 3 and 5 years to the respective period from the Mercer MPA Australian Shares universe as at January 2009.
1 Gallagher, D.R. and Gardner, P., (2005), “The implications of blending specialist active equity management”, University of NSW, School of Banking and Finance
2 Eggins, J, “Understanding the ingredients in the multi-manager portfolio pot”, JASSA Issue 2, 2007
3 Gallagher, D.R. and Gardner, P., (2005), ‘Portfolio Design and Challenges Inherent in Multiple Manager Structures’, JASSA (The Journal of Finsia — Financial Services Institute of Australasia), Summer(4), pp 20–25.
4 Brands, S. and Gallagher, D.R. (2004), “Risk and Return properties of fund-of-funds”. JASSA Issue 1, Autumn 2004
5 Eggins, J, and Parish, S. “Two much of a Good Thing? The fallacy of over diversification” Russell, March 2007.
Opinions, estimates and projections in this article constitute the current judgement of the author as of the date of this article. They do not necessarily reflect the opinions of Schroder Investment Management Australia Limited, ABN 22 000 443 274, AFS Licence 226473 ("Schroders") or any member of the Schroders Group and are subject to change without notice. In preparing this document, we have relied upon and assumed, without independent verification, the accuracy and completeness of all information available from public sources or which was otherwise reviewed by us. Schroders does not give any warranty as to the accuracy, reliability or completeness of information which is contained in this article. Except insofar as liability under any statute cannot be excluded, Schroders and its directors, employees, consultants or any company in the Schroders Group do not accept any liability (whether arising in contract, in tort or negligence or otherwise) for any error or omission in this article or for any resulting loss or damage (whether direct, indirect, consequential or otherwise) suffered by the recipient of this article or any other person. This document does not contain, and should not be relied on as containing any investment, accounting, legal or tax advice. Schroders may record and monitor telephone calls for security, training and compliance purposes.