Multi-Period Performance Attribution: A Refreshing Perspective

Abstract:     
We start by reviewing the basics of the single-period Brinson and Fachler (1985) model, which differentiates between an allocation and a selection effects when explaining the active return of a money manager. Contrarily to single-period measurements, there is no undisputed multiple periods attribution methodology. We present some of the most popular algorithms used in the performance measurement industry and discuss some properties from an entirely new perspective. We first demonstrate that none of the algorithms under analysis conforms to the natural laws of causality and homogeneity of time. Second, we use Monte Carlo simulations to analyse the impact of the revisions incurred by the Frongello (2002) multi-period linking algorithm on the factual single-period attribution effects. We demonstrate that the impact of the revisions is large on average and highly volatile. The basic arithmetic and geometric calculations seem therefore more appropriate. We finally find two regimes that discriminate between a proper use of either the arithmetic calculation or the geometric compounding.

Number of Pages in PDF File: 7

Keywords: Performance attribution, selection and allocation effects, multi-period linking, Monte Carlo simulations

JEL Classification: G10, G11

working papers series