Apart from FRS17, the two big events in 2001 for UK pension funds were the Myners Report in March and the Unilever/Merrill trial from October to December. If they shared one thing in common, it was that they referred extensively to the way we measure performance of pension funds, and more specifically to the move away from peer group towards scheme-specific benchmarks.
Myners gave the transition his whole-hearted support, but the legal case showed just how badly things can go wrong when that transition gives rise to misunderstandings between the client and manager. There are many lessons to be drawn from the case, but with a third or more of UK funds still to make the transition, the unfortunate experiences in the Unilever/Merrill case surely deserve attention.
Indeed, the message of this short article is that setting objectives and benchmarks, and how these inform and guide managers’ behaviour, has relevance to all those in the pensions industry, both clients and managers. Furthermore, this applies equally to defined benefit and to defined contribution schemes.
The Unilever Superannuation Fund was certainly not the first to make the move to a scheme-specific benchmark. When it did so in 1996, roughly 5% of all funds had made the move every year since 1990.Unilever was by no means a pioneer. Yet five years later, in 2001 about a third of all UK schemes had still not made the switch. What is it about these laggards that so irritates our thought leaders? For Myners’ is not a lone voice. In brief, it is that a peer group benchmark arrangement conflates the two big tasks facing the stewards of our pension assets: setting long-term policy on the one hand, and implementing it on the other. Others call the first the governance task, and the second the execution task. Strategy and tactics, long-term and short-term, are further word-pairs. In a peer-group benchmark world, the investment manager does both jobs, and the fox
is very definitely in charge of the chicken coop.
In such a world, managers often adopt a ‘one size fits all’ approach. There is little, or no, differentiation between client portfolios to reflect differences in the quality of the sponsor’s covenant, or in the demographics of each scheme. If there is a differentiation, it is between the +1%-ers (over the peer group median) and the +2%-ers. However, as schemes move to tailored benchmarks, the expression of the individual goals and risk tolerances of each client take on new significance, and they can be more precisely tailored. Of course, there is a corresponding increase in the monitoring burden within investment management firms, for no two funds are necessarily alike in the expression of their requirements.
The original objectives of the Unilever scheme called for Merrill and its other managers to “maximise long-term returns subject to an acceptable level of risk normally associated with a balanced approach to fund management”. This really was a ‘managers know best’ approach. By contrast, the new benchmark had a target rate of return of the benchmark plus 1%, as measured over three years, and a downside risk tolerance of –3%, as measured over any trailing four calendar quarter period. Effectively, the more precise wording of the new objectives, and the specification of a benchmark portfolio, reflected the greater responsibility taken by the client for setting the scheme’s investment policy. Indeed, the Pension Act of 1995 seems to demand that trustees, not managers, take on this responsibility.
In the accompanying diagram, the left hand block represents the activity of defining policy. Policy is set, more often than not, in the context of an asset/liability modelling exercise. In the Unilever case this was entrusted to Towers Perrin in 1995. The ALM study was part of a review that led to the new benchmarks. In Unilever’s eyes, the transition had been expensive in management time and money. There were bills for actuarial advice, and bills for investment manager selection advice. There was a multi-million pound bill for the portfolio transition management to implement the new manager configuration. If this effort and expense were to yield fruit, then the managers must play their part by staying within the new guidelines.
On the right hand side of the diagram are the investment manager’s core responsibilities. If the arrangements call for active management, then the first task is to identify outperforming asset classes and securities. It was clear in the Unilever /Merrill case, that the judgments made by the manager in charge of the portfolio were extremely disappointing as measured by the relative performance of his favoured holdings in 1997. But this was common ground. Merrill – or rather Mercury, as the firm was then named – was not under attack from its former client for those judgments, or for the quality and diligence of the research that underlay them. The dispute was about the next box down – portfolio construction. According to barrister Jonathan Sumption, in his opening address to the court on behalf of Unilever, the manager concerned “… constructed a portfolio for my client whose performance was sensitive to an extreme degree to the correctness of his judgments about the market”.
In our diagram, the investment management agreement, or mandate, sits as a nexus between these governance and execution roles. Like a fulcrum, it bears a considerable weight, and therefore needs to be treated with great care on both sides of the governance/execution divide. The portfolio construction task is guided not only by the appraisal of securities and markets, but also by careful attention to individual – even idiosyncratic – expressions of the client’s goals and risk tolerances, as encapsulated in the mandate. These are the critical inputs to the portfolio construction task.
Merrill’s defence was double-barrelled: they argued that the portfolio’s construction, and therefore its risk characteristics, had been checked using the BARRA model – an industry standard risk model. They also said that, in any case, portfolio construction is more of an art than a science. On the use of the BARRA model, the court case focussed on the limitations of quantitative models, and on the dangers of excessive reliance on them. They work fine, as long as the future to which they address themselves bears a passing resemblance to the past on which they are based; unfortunately this is all too often not the case. Also, they often incorporate assumptions that may, or may not, be upheld; for example, the annualised tracking error is, in fact, a monthly estimate multiplied by the square root of 12. That is unsafe if there is persistence, ie positive serial correlation of relative returns. In the Unilever/Merrill case, performance was negative for eight quarters in a row, from the second quarter of 1996, to the first quarter of 1998. Analysis of Merrill’s past performance record showed that it had indeed been ‘fat-tailed’ ie non-random, since inception in 1987.
The court also dwelt at some length on the significance of the asymmetrical horizons for upside performance – three years, and for downside performance tolerance – four trailing quarters. To see the significance of this, imagine that you budget for speeding fines on the basis of a known number of roadside cameras and a given driving style. What happens if more – four times as many – cameras are installed? Something has to give, either your risk budget or your driving style. The claimant’s side argued that the ex-ante tracking error consistent with the new mandate was closer to 2% than the 3.5% and higher that the manager initially adopted.
The paradox of the Unilever/Merrill case is that it simultaneously brought home the limitations of quantitative models, but also the importance of their intelligent application. Performance measurement produces apparently hard numbers, but if the underlying stochastic processes that produce them are themselves unstable, then how should we interpret the results? Our diagram places performance measurement and attribution as an important part, but not the totality, of risk management. The overseeing of these processes requires good numeracy, but also the ‘softer’ interpretative skills of seeing the big picture. It requires that each party to the contract understands not just its own role, but also of what the other party is responsible for. The trial showed that the task of clear (and continuing) communication between manager and client was perhaps bigger, more necessary, and ‘trickier’ than even experienced managers and trustees had believed – especially in a time of transition.
Mark Tapley acted as an adviser to Unilever in its recent court case against Merrill Lynch in London

Topics