Decline of core-satellite approach
The pension fund community in the US is seeing the beginnings of a paradigm shift in the way plan sponsors think about asset allocation issues. Disillusionment is growing with a couple of the key drivers behind most strategic asset allocation and manager hiring processes over the last 10 to 15 years.
The first of these is the role of a strategic asset allocation benchmark. When times were good, it was easy to forget that the ultimate purpose of a defined benefit (DB) fund, for example, is to match the liabilities of the fund at the lowest cost to the plan sponsor. This means that the natural benchmark for the fund is the liabilities themselves. Of course, in the course of the standard asset-liability modeling that takes place, the strategic benchmark will hopefully to some extent reflect current liabilities but cannot of course reflect changes in those liabilities. Perhaps more importantly, the asset liability modeling exercise itself was not typically flexible enough to suggest the structural aspects of the allocation that would best enable a close match to liabilities. None of this, of course, matters very much when equity markets are going through the roof. But it matters now.
The second is disillusionment with the style and cap distinctions that have led to the multiplication of benchmarks for managers to manage against. These distinctions have greatly complicated management issues for plan sponsors, taking up time in searches, manager reviews and so forth as well as costing a lot in the form of active fees.
The first issue can easily be addressed. The right approach to strategic asset allocation explicitly takes the liabilities into account not just as part of a three year or five year cycle of full scale reviews, but as the default benchmark against which the funds performance is judged on a month by month basis. This helps in a number of ways, most importantly, the plan sponsor can see what is happening to their surplus (or lack of) on a much more regular basis and if needed, make adjustments to the balance between liability matching and return seeking instruments in the portfolio. It also becomes easier to see the structural problems in the plan caused by concentrating too much on strategic asset benchmarks. For example, in the US the overwhelming majority of domestic fixed-income is benchmarked against the Lehman Aggregate index. This is a perfectly fine benchmark but as of February this year, had a modified duration of just under four years. Most funds will have liability durations up in double digits so that using the Lehman Aggregate as a benchmark, builds a structural and substantial short duration bet into the portfolio. Now you might be comfortable being short duration relative to the liabilities or you might not, but in most pension fund systems you won’t ever see the impact of that bet as part of the attribution of performance, let alone have a clear idea who is managing it. Direct management relative to the liabilities makes the bet stand out like a sore thumb and just as importantly, if the asset classes used in the optimisation relative to liabilities include long-duration instruments and inflation indexed instruments, typically these will find a role in the portfolio as both better liability matches and useful ways of dampening the volatility of the surplus.
We have been able to recommend specific improvements to the way in which fixed-income is used in DB plans in the US based on direct management of the surplus utilising inflation indexed and overall fixed-income duration tailored to a fund’s specific liability duration.
The second important issue is the role of style and capitalisation boxes for the equity portion of the allocation. The case for dividing up equity allocations into style-cap boxes is that it permits better diversification and that there are identifiable premia attached to the various categories. However, whether style boxes really permit better risk control is very much an open question. Multiplying the number of managers reduces the ability of the fund to manage the overall risk of the plan and indeed ‘throws back’ decisions to the plan sponsor that they might reasonably expect their managers to take. If growth outperforms value for two or three years, should the plan sponsor be rebalancing or letting the growth portion of the fund grow as a percentage of the total allocation? Value/growth timing is a tough call, why put in place a structure that forces the plan sponsor to make such a call? This happens, fundamentally, because managers can only manage their risk relative to the benchmarks they are asked to manage against rather than the underlying risk of the plan itself.
Multiple managers complicate the risk management of the fund in another way. Table 1 shows an example where the plan sponsor wants to take 200 basis points of active risk in the equity portion of her portfolio. Column two shows what happens to the aggregate tracking error assuming that excess returns are uncorrelated across managers and that each manager is given a tracking error target of 200 basis points. The aggregate tracking error falls as more managers are added so that with four managers the aggregate comes out at 1% tracking error. At this point the plan sponsor is paying active fees for at best enhanced index performance. Now a smart plan sponsor can of course adjust the target for each manager so that in aggregate the predicted tracking error is 200 basis points. With four managers, given our assumptions, the target for each manager needs to be 400 basis points. This works as long as the assumptions hold. However, the assumptions tend to fail just when you need them the most. When markets are volatile, return correlations within a given asset class tend to rise. This is one reason for excess return correlations to increase. At the same time, typically as markets approach extremes, the positions the managers take also tend to become more highly correlated. So, for example, if we consider growth managers in 1999, many would have realised the internet bubble was looking ever more likely to pop. Thus, they would have tended to underweight this sector and their overweights would also therefore become more correlated. With both the underlying returns and the manager positions becoming more correlated at the same time as market volatility was increasing, the end result is that the apparent diversification benefits of multiple managers disappeared and more so that the aggregate risk in our four manager example, approaches and actually goes through the 4% level as a result of much higher excess return correlations, higher manager position correlations and increased market volatility. This is illustrated in chart 1, which shows the changing excess return and volatility patterns of five randomly selected large cap growth managers.
Using divisions like value and growth or large and small cap only really makes sense for a plan sponsor if there are good reasons to believe that a manager needs very specific skills to excel in each area. In the case of value and growth this is an extremely hard case to make. Stocks often move between one index and the other, sometimes multiple times over the years. This leaves one arguing that a manager may have had specific expertise in this stock last month but that now they do not, but of course they may well have insight into this stock again after the next reconstitution of the index. This is the pragmatic version of the statistical analysis which suggests that there is no stable value-growth premium, or stable small cap premium.
An alternative that is gaining ground in the US, paradoxically as style fever seems to be moving to Europe, is a much simpler framework. In this simpler framework, passive equity and enhanced indexing form the bedrock of the equity portfolio. The more active component is made up of a small number of active managers with broad benchmarks who are chosen not because they fit a box but because they are credible and fill a need in the portfolio. They might be managers with a style bias, that they might not be the key here is that is not the primary determinant of whether or not they belong in the asset mix.
This is intriguing because it is a cheaper structure in terms of fees and one that has fewer arbitrary assumptions about style built into it. It also helps address the first issue we tackled, managing a fund’s assets more in line with the evolution of the liabilities. This is much easier to do with a simpler overall plan structure. By freeing up the plan sponsor from dealing with multiple managers in a whole array of style boxes, the greatest virtue of this simpler approach may be that it saves that most precious commodity: time. Allowing the plan sponsor opportunity to focus on the bigger issues facing the plan. In the US, style is finally going out of fashion. In Europe, the right place for the style police is in Armani or Paul Smith, not in your portfolio.
Tony Foley is director of the Advanced Research Center at State Street Global Advisors in Boston