Trustees face a bewildering range of performance and risk measurement tools, presented as essential aids to managing a successful pension fund. These include performance measurement algorithms and attribution techniques, performance presentation standards and various methods of measuring and evaluating the risk undertaken.
Why do we need to devote effort and resources to understanding past performance down to the last basis point? This article argues that, used correctly, analysis of historic performance has a value and that using these techniques within a pension fund environment is easier than might be expected.
Funds frequently measure success simply by comparing with a median of other supposedly similarly managed funds. These funds may have different benchmarks, making the comparison irrelevant. It is important to be able to measure whether the objectives of a fund have been met. Too often they are stated in pithy terms such as “to achieve maximum return at low risk” or “to be in the first quartile”. This provides little insight into how effective trustees have been as decision-makers and how much value has been added by the selection of managers.
The key is to understand and evaluate in terms of the individual decisions made in managing the fund. Performance analysis can be a useful aid to this; outstanding asset allocation decisions can easily be overshadowed by the wrong choice of managers or investment styles within an asset class. In turn, a stellar performance by managers versus their allocated benchmark may mask some major problems in top-level asset allocation.
A recent survey by William M Mercer for Lend Lease Corp identified poor implementation procedures as a source of under-performance. These specifically included unplanned exposure to style biases and long delays in firing managers with organisational problems. A clear logical analysis of past performance could well have identified these before they became a consistent drag on performance.
We first need to ask some obvious but very important questions:
q What are the key objectives and needs of the fund?
q What are the main decisions made to achieve these objectives?
q How does responsibility for these decisions break down between trustees and the appointed investment managers?
q How is the success of decisions made by these parties measured?
Not all pension funds have the same decision-making process. However, most will make broadly the same key decisions. Specifically, how to structure the fund (using balanced, specialist or index managers), which managers to choose and what mandate to give each manager. It is easy to fall into technical discussions on algorithms to use; more productive is to clarify the investment objectives, break down the decision-making process and identify responsibilities for these decisions. The portfolio needs first to be broken down into major asset classes to reflect the top-level asset allocation decisions by the trustees. Selected managers within an asset class are then assigned a benchmark as their starting point for decision-making and a measure of their success. This approach will allow the attribution to clearly reflect decisions on asset allocation (mainly trustee responsibility) and stock selection (mainly manager responsibility).
A further level of analysis could compare the performance of a manager against his assigned benchmark; this would break out where value was added through his allocation decisions and through pure selection. This can help to identify problems in an organisation or potential for success.
Figure 1 illustrates this layered approach, enabling explanation of under- or out- performance at each decision level. We look now at a typical UK pension fund and apply the analysis. The fund structure and benchmarks at each level are identified in figure 2. Key allocation decisions made by the fund described above are identified in figure 3.
At the top level we can analyse the impact of the asset allocation decisions, for example, the decision to underweight exposure to the UK and property while over-weighting Europe, the US and Pacific Basin.
We may then probe one level deeper and analyse whether the UK balanced manager made a correct decision to underweight UK equity and overweight UK fixed income relative to his assigned benchmark. Decisions to allocate between active and passive management within an asset class may also be evaluated at this level.
We now turn to the specialist managers and look to reasons for their under- or out-performance relative to their benchmark. Did our Pacific Basin manager overweight Japan during recent stock market turmoil? Or did he make poor stock selection decisions within a particular Pacific market? Does this give us any indications of a serious problem within the organisation or was it just a particularly difficult period?
Once the basic framework is in place, we can focus our attention on the specific results that have the most impact for the fund and its management. The analysis should enable us to pinpoint where problems and strengths lie and should be seen as a starting point for action and follow up
We can extend this to other decisions and use other analytical tools such as risk analysis and portfolio profiling within the same decision framework. This is particularly important for risk analysis where each manager may have an assigned risk tolerance. However, total portfolio risk is not a straight weighted average of these, due to potential benefits of diversification across styles and markets. We need to attribute any such benefits to a successful asset allocation policy rather than to any particular manager.
Turning to the practical implementation of this analysis process, the immediate reaction may be one of horror at the amount of data required and the potential cost in obtaining such data. This data already exists in the form of valuations and transaction reports provided by the custodian. To avoid resources being spent on data mining rather than on analysis, the most efficient route must surely be to utilise data direct from the custodian system. There have been significant moves forward over recent years, both by custodians in developing open access to data, and by the providers of performance and risk analysis in developing interfaces to custodian systems. Integration of data means attention can be focused on using and acting on the analysis rather than on data requirements.
As alluded to earlier, we are told not to rely too much on past performance when hiring a new manager or assessing the likelihood of future success. We should not ignore the need for strong leadership and a commitment by the investment team to a rigorous and consistent process. Performance analysis alone will not reveal this for any particular firm but it can act as a valuable check, questions must be asked of a manager after four years of fourth quartile performance. Used within the correct decision-making framework, attribution and risk techniques will identify how successful individual decisions have been, and will highlight sources of risk and re-turn in the portfolio. A clear structure for analysis means these techniques do not need to remain solely in the domain of rocket science!
Suzanne Mitchell is assistant manager, performance measurement, at Russell/Mellon Analytical Services