Performance Attribution: Introduction

[this page | pdf | back links]

The Nematrian website differentiates between fund-level performance measurement calculations, explained in pages linked to PerfMeasIntro, and stock, sector and factor level performance attribution calculations described in these pages.


Employers of investment managers will generally be interested in understanding what has been the source of their (past) under- or out-performance. The process of carrying out such an analysis is typically called performance attribution.


Performance attribution typically requires high quality accounting and/or instrument characteristic data. However, it may not necessarily need as high quality data as may be needed for pure performance measurement purposes.


From a theoretical perspective this differentiation arises principally because sources of under or outperformance interact with each other. It is therefore not always practical to expect performance allocation to provide a unique theoretically correct answer. For example, suppose a manager performs well in an asset class that also did well and in which he/she was overweighted. Should we deem the additional outperformance at the fund level coming from the good performance on the overweighted asset class be deemed to be good ‘stock selection’ within that asset class, or good ‘asset allocation’ between classes? Once users focus on such uncertainties they may also feel that the need for high precision is less compelling, particularly if it comes at a high cost.


A particular issue here is that some types of transactional data we might otherwise ideally want, i.e. prices at which instruments are bought and sold, can in theory be dispensed with for performance attribution purposes if we include in an ‘other’ category the contribution to performance coming from the fund manager buying and selling instruments at other than their period end valuations.


However, the problem with this line of reasoning is that we can always identify situations where such blurring does not actually apply in practice. In the situation highlighted in the previous paragraph this would include the case where the fund manager did not deviate materially from benchmark exposures but typically bought assets at a poor price. The element of the attribution that we would then ideally want to focus on would have been bucketed into the residual item that had been created by lack of better data.


Contents | Prev

Desktop view | Switch to Mobile