Fund Labeling Tests

I test fund labeling effectiveness using 276,317 fund/month observations over the 1988-2007 time period by estimating two cluster analysis goodness-of-fit metrics: 1) across cluster correlations and 2) skill returns. Lower correlations and higher returns are better. The basis for this cluster analysis goodness-of-fit test is comparable to risk-return optimizing an investment portfolio.

The results presented in the Graph below reveal that Style Box labeling is inferior to all other approaches, including Base clustering, which assigns funds at random. The preferred direction in the Graph is northwest, which means lower correlations and higher returns. However, Style Box labeling moves in the opposite direction, with higher correlations and lower returns. Since Style Box labeling plots below the random assignment Base regression line in the Graph, it performs worse than does random labeling. This means that Style Box labeling produces heterogeneous strategy clusters and thus makes it difficult to identify top performing managers within a particular strategy.

Both Strategy labeling and Strategy- successful active manager (which is labeled a Diamond Rating 5 manager or DR5 for short) labeling are superior, with Strat-DR5 performing as well as one can expect. This is demonstrated in the Graph by the fact that only 4 of the 1000 random fund assignment Base simulation risk-return pairs are better than that produced by Strat-DR5 labeling.

Thus Strategy labeling moves in the right direction, producing lower correlations along with higher returns. This means that Strategy labeling leads to the formation of homogeneous strategy clusters and allows for easier identification of top performing managers within each strategy. Strategy labeling is based on the information contained in the “Principal Strategies” section of the fund’s prospectus and so these results support the usefulness of this information for fund labeling.

Conclusion

Strategy labeling is very effective at clustering funds pursuing the same strategy and at identifying skilled active managers within these strategies. On the contrary, Style Box labeling is very ineffective since it spreads common strategies across clusters rather than concentrating them within clusters.

Both Style Box and Strategy labeling have their problems, as Style Box is based on the premise there is a strong time consistent link between strategy and the characteristics of the stocks held by the manager, while Strategy is based on the reliability of information provided by the manager and fund. These issues can only be addressed by empirically testing which approach is best at clustering funds pursuing the same strategy and allowing for the identification of skilled managers within these strategies. The evidence strongly supports Strategy labeling as being the more effective approach.

Copyright 2008 by AthenaInvest, Inc. All rights reserved.
Do not quote without permission.

Display article as PDF for printing.

Would you like to send this article to a friend?