If the performance of a fund or portfolio is different from that of its benchmark, what explains that difference? This key question is regularly asked by all stakeholders in the asset management process: those who manage investments in funds or mandates, the marketers of investment products, and those who select and use those products.
So, a pressing question with a big potential audience – but, amazingly, not one that can be answered in most parts of the investment world.
In this piece we track a conversation between three Apex Risk Technologies principals – Alessandro Maimone, Diego Martinez, and Magnus Spence – who discuss why this question is important, and to whom, and why it is not being answered. The conclusion they reach: if your data is well-managed (and we addressed this issue in our last blog) there is no reason why you cannot provide a reasonable answer, reasonably quickly, and very cost effectively.
Martinez starts by setting the scene: “An initial question that we find is often asked about investment performance is: ‘How is return distributed across asset classes (or sectors or region or other factors)?’ And the answer to this important question is provided by Contribution analysis. This measure is useful for funds not managed to a benchmark such as Long-Short portfolios and some Multi-Asset funds.”
Maimone continues: “But I think we would agree that the performance question that we want to focus on in this conversation is this: ‘If the performance of a fund or portfolio is different from that of its benchmark, what explains that difference?’ Is excess performance driven more by the choice of asset classes or regions (or other factors), or is it the choice of individual securities within those asset classes (or other factors), or is it something else? Attribution analysis gives the quantified answer to this question. This question is asked by funds managed to a benchmark which covers traditional long-only and equity funds.”
Spence adds some history. “Attribution analysis can be tracked back to the 1960s when terms like asset allocation and security selection were first used in assessing performance of large pools of pensions assets. The US asset manager Gary Brinson (who later went on to found Brinson Partners and eventually to be the CIO of UBS Global Asset Management) and others took that early work and evolved it 40 years ago into a measure of excess performance relative to a benchmark, that is now referred to as the Brinson Attribution method .”
“The Brinson method” develops Martinez, “has been the most popular attribution approach since then – it calculates the contribution to excess return from each of three elements: asset allocation, stock selection, and ‘interaction’ .”
“We need to address the still often quoted ‘fact’ that asset allocation determines 94% of performance” suggests Spence. “This finding is derived from a paper written by Brinson and others in 1986. There has been much work done since then to revise this conclusion. It is now generally agreed that ‘asset allocation and security selection are likely to be equally important, depending, of course, on the investment approach taken.’ ”
“The Brinson model effectively tells you which decisions have worked and why, in a way that no other measure can really do”, says Maimone. “Knowing – for example - that it is the decisions on a portfolio’s security selection that have driven performance, rather than the decisions on portfolio rebalancing around which sectors investments should be reallocated to, will guide the thinking of managers of the fund and how they act in future.”
The benefits go even wider. The Brinson method helps shareholders/clients in the fund to form views on the skills of the managers by explaining both under and over performance and it will help investment firm senior management to assess the success of their colleagues, and their adherence to investment guidelines.
Given all the potential advantages for such a wide set of stakeholders in having attribution measures to hand on a regular basis, why is it that Attribution analysis is carried out by only limited numbers of investment industry actors?
Spence adds some colour: “For example, in the UK Pensions industry, it is still – even after 40 years - rare for schemes to be given a detailed attribution analysis, and many are shown no such analysis at all. Only the very largest institutional investors are likely to have a full attribution analysis carried out. It is not common practise for asset management firms to provide attribution analysis of their own funds for either their institutional or retail clients. And wealth management / private banking firms do not provide this form of analysis for their individual clients. So - significant elements of the investment world have had to rely on inadequate investment performance analysis for nearly 40 years”.
Why is this valuable and apparently straightforward measure not more widely used?
“Well, we have to accept that this analysis is not the answer to everything,” concedes Martinez. “The neat simplicity of the Brinson model is admirable, but many will argue that it has to be adapted to deal with measurement over multiple periods, and multiple currencies, and it loses its simplicity in so doing. It doesn’t suit all investment portfolios: fixed Income excess performance is less easy to analyse using the model, and Illiquid/unquoted investment analysis does not work well.”
There are also technical data-related obstacles, says Maimone. “Brinson analysis requires regular and consistently formatted data over time. Well-structured data management can be a challenge to some asset managers as we explained in our last blog. Providers of Benchmarks frequently do not provide data in a format that is appropriate for this purpose, which adds to the challenge”.
“Cost is an issue too” accepts Martinez. “The analysis requires someone to have access to (and thus pay for) the appropriate benchmarks per portfolio. This cost can be a barrier for even for large firms and certainly puts off smaller pools of assets even including sizeable pension funds”.
All agree that even after recognising some of these barriers to its use, Attribution analysis deserves much wider acceptance.
Apex Risk Technologies can carry out this analysis and embed it into risk governance within investment firms, quickly and easily as part of a suite of risk management measurements.
“There really is no excuse for avoiding this valuable and easily generated 40-year-old measurement any longer.” concludes Maimone.
Contact us for more information
- Strictly speaking it is the Brinson-Fachler method developed in 1985 that is most used today.
- The interaction effect is the combination of the selection and allocation effect. If the portfolio allocation outweighs and outperforms the benchmark, the interaction effect is positive, and vice versa. Apex Risk Technologies interprets interaction effect as the part of active return that is not directly attributable to selection and allocation. Some (such as Morningstar) incorporate interaction effect into either the allocation or the selection effect, depending on the investment approach taken by the investment manager.
- The best-known revision came in 2001: “Does Asset-Allocation Policy Explain 40 Percent, 90 Percent, or 100 Percent of Performance?” (Ibbotson and Kaplan). A 2021 example of the revisionist view is “Asset Allocation: From Theory to Practice and Beyond” (Kinlaw, Kritzman and Turkington). The quotation used here comes from the review of this book by the CFA institute Book Review: Asset Allocation | CFA Institute Enterprising Investor