The growth of multimanager portfolios has highlighted the importance of the ability to rank and select investment funds. Traditionally, the Sharpe ratio has been used as a risk-adjusted performance measure in ranking managers. The selection of top funds can be improved by using clusters or groups of performance measures instead of just the Sharpe ratio.
What’s Inside?
An examination of European equity funds demonstrates that fund performance is structured around three dimensions: risk-adjusted return (backed by the Sharpe ratio), active management skill (supported by the information ratio), and market sensitivity (beta). These components explain about 70% of the variation in performance. The author uses clustering techniques to support fund selection according to a particular performance dimension. Additionally, a rank-correlation analysis and a back-testing exercise are conducted to ensure robustness. The author finds that the three-year Sharpe ratio has advantages over other traditional measures. But when the Sharpe ratio is used alone, instead of as part of a cluster group, it is not as reliable.
How Is This Research Useful to Practitioners?
Standard performance measures, such as the Sharpe ratio, the information ratio, and beta, are often used when ranking funds. But clustering data-mining techniques appears to provide a better solution for the analyst. Clustering techniques are useful for managers in that they allow the comparison of performance measures. Analysts can use cluster structures to create a flexible approach to analyze performance. Clustering also allows an investor to select funds according to a preferred performance component.
An integrated approach that includes all performance measures, regardless of the period of calculation, leads to more clear-cut clustering groups. Cluster-scoring coefficients can be used for comparison between funds, possibly enabling investors to find arbitrage opportunities.
How Did the Author Conduct This Research?
European equity funds from the Reuters–Lipper database for the period of 2000–2007 are used for the analysis. After eliminating highly correlated measures, the author uses nine performance measures for each fund over a six-month, one-year, and three-year period. They include three traditional risk-adjusted performance (RAP) measures, four performance measures obtained from regression, and two measures related to benchmarks. Cluster data-mining techniques are used to group these performance measures into eight optimal cluster classes.
The author demonstrates the advantages of clustering by performing a rank correlation analysis and a back-testing exercise. The rank correlation results show that different fund selections would occur on the basis of the chosen cluster as well as the Sharpe ratio rankings. As a result of this outcome, the author examines which cluster would allow for the best selection of funds. Using a back-testing exercise, the author finds that the combination of Jensen’s alpha, absolute performance, and Sharpe ratio offers a superior ranking for fund selection.
Abstractor’s Viewpoint
Given the prevalence of available performance measures, a method to compare performance measurements is valuable. Through clustering processes, the author demonstrates an effective method to analyze performance and aid in fund selection. But one problem with the article is that the dataset only covers a pre-2008 time period. If the data were extended into a post-2008 time period, when there were negative returns, the Sharpe ratio would have some problems when ranking performance.