Issue: September/November 09


Valuable tools

Mulder...useful data

Mulder...useful data

Ever tried to understand what PCA and PAA are about? You should! SUNETTE MULDER, senior policy advisor at the Association for Savings & Investment (Asisa), explains all.

Due diligence and client reports contain a review of retirement funds’ portfolio performance. Although this review might seem straightforward, it is often misunderstood.

Two concepts that trustees come across are performance contribution analysis (PCA) and performance attribution analysis (PAA). The terms are commonly used interchangeably. Technically, however, they refer to different sets of information that trustees should be able to understand.

Performance contribution analysis

PCA breaks down, into components, the return achieved. This can be shown on asset class, sector or stock level. It is calculated by multiplying each component’s weight and return, representing the particular component’s contribution to the total return of the fund or benchmark. The sum of each contribution will add up to the fund’s total return.

For the trustee or consultant, PCA can be valuable in assessing the capability of an asset manager to manage assets in a particular style and will highlight the biggest positions taken. This should stimulate discussion with the manager on strategy.

Even more analysis is possible when a fund’s contribution is compared with a benchmark’s contribution. However, at this level one is not yet looking at relative out/underperformance, but on an absolute-return level, i.e. what contributed to a fund’s return of x% and what contributed to the benchmark return of y%.

A potential issue in calculation is that, for periods of longer than a month, the weights used represent an average over the period under review. This can distort the results, especially for periods of a year or longer.

Performance attribution analysis

PAA aims to explain relative over/underperformance. So an attribution is always run against a benchmark. If a fund does not have an investable benchmark (e.g. absolute-return fund with a cpi benchmark), a performance attribution cannot be done except if a proxy or attribution benchmark is used. Rather, a performance contribution should then be done.

An attribution will break down excess performance into the three attribution effects of allocation, selection and interaction. The latter does not form part of the investment decision as it relates to the combined effect of asset-allocation bets with stock-selection decisions.

The general market practice is therefore to combine the interaction and selection effects. The sum of the three effects adds up to the active return or out/ underperformance generated on a particular fund. When an attribution analysis is done on stock level, for purpose of this article it relates to equity attribution and not fixed-income attribution.

The allocation effect for each asset class/sector/ stock indicates the value added/detracted by an over or underweight position in a good or bad performing asset class/sector/stock. The overweight or underweight position is calculated on an average weighting during the period and the performance of each component is measured relative to the total return of the overall benchmark.

When an asset manager decides to hold a stock that is not a component of the index, it is still an active decision and should be included in the analysis. Likewise, the decision to not hold a stock that is in the index should be included in the analysis.

When the effect of selection including interaction is analysed, it’s important to distinguish between an analysis done on an asset class/sector level or on a stock level.

At the asset class/sector level, the selection (including interaction) effect indicates the value added/detracted by selecting particular holdings within an asset class/sector. In other words, for a particular weight of an asset class/sector, did the asset manager manage to out/underperform the benchmark’s asset class/sector by selecting the right or wrong securities?

The selection (including interaction) effect on a stock level intuitively seems to indicate good or bad stock selection. In fact, however, it indicates whether the asset manager is generating a higher/lower return in a particular stock by trading at the right/wrong time. So, at this level, it should be seen as a trading effect.

PAA is a key tool to understand a fund’s sources of return against its benchmark. It serves as a starting point for the asset manager to discuss the positive and negative aspects of a fund’s recent performance. It also gives the trustee or consultant an indication of whether the asset manager excels at asset allocation or stock selection. This analysis highlights whether the positions taken relative to benchmarks paid off.

Although possible, PAA is generally not calculated for periods longer than a year as the effects generally are distorted by the average weights being used. This problem can be minimised by the asset manager calculating attributions on a monthly basis, then linking these monthly attributions to minimise the potential error.

Fixed-income attribution is not done on a similar basis as the investment strategies employed by bond managers differ significantly from equity managers. Models developed for equity attribution are unsuitable.

Trustees and consultants should know whether the approach of the investment manager is based on top-down or bottom-up decisions. Top-down is when stock selections flow from the asset-class and/or sector-level allocations; bottom-up is when asset-class or sector-level allocations flow from stock selections. Clarity can help informed discussion between asset manager and client, and lead to better understanding of the investment process being followed.

Some asset managers might show the return achieved by a fund’s different asset classes against relevant benchmarks under a ‘Performance attribution’ heading. Technically, this is merely a comparison of returns. If necessary, the trustee or consultant should request additional information from the manager.

PCA and PAA are in reporting. Trustees and consultants should understand the tools’ respective aims, and gain value from them both.