Performance Attribution and the Accuracy of Detecting Timing and Selection Skills

Abstract

In this study we investigate the accuracy of the popular Brinson, Hood, and Beebower [1986] attribution model in identifying investors with forecasting abilities. The purpose of the Brinson, Hood, and Beebower model is to classify the forecasting abilities of a portfolio manager into timing and selection skills. In our study, we performed two simulation experiments, one with managers having only timing skills and one with managers having only selection skills. Furthermore, we model two different types of portfolio managers. The first manager is able to forecast the direction of the future stock returns relative to a benchmark, whereas the second manager is also able to predict a part of the magnitude of the future performance. Our results suggest that the measurements of timing ability are associated with a lot of uncertainty. This is rather surprising, since the model uses also direct observations of the portfolio composition in the estimation of the timing ability. On the other hand, selection ability can be detected with far better accuracy. Furthermore, we conclude that the nature of the forecasting ability has an impact on the reliability of the evaluations.

Keywords: Performance measurement, forecasting abilities, performance attribution, timing, selection

JEL Classification: G14, G23, G12