On the informativeness of persistence for evaluating mutual fund performance using partial frontiers

• We apply FDH, order-m   and order-αα partial frontiers to a sample of US mutual funds.
• We merge these methods with the literature on mutual fund performance persistence.
• An algorithm relating m   and αα to performance persistence and the contrarian effect is derived.

The last few years have witnessed a rapid evolution in the literature evaluating mutual fund performance using frontier techniques. The instruments applied, mostly DEA (Data Envelopment Analysis) and, to a lesser extent, FDH (Free Disposal Hull), are able to encompass several dimensions of performance, but they also have some disadvantages that might be preventing a wider acceptance. The recently developed order-m and order-αα partial frontiers overcome some of the disadvantages (they are robust with respect to extreme values and noise, and do not suffer from the well-known curse of dimensionality) while keeping the main virtues of DEA and FDH (they are fully nonparametric). In this article we apply not only the non-convex counterpart of DEA (FDH) but also order-m and order-αα partial frontiers to a sample of US mutual funds. The results obtained for both order-m and order-αα are useful, since a full ranking of the mutual funds' performance can be obtained. We merge these methods with the literature on mutual fund performance persistence. By combining the two literatures we derive an algorithm which establishes how the choice of m   and αα parameters intrinsic to order-m and order-αα (respectively) relate to the existence of performance persistence and the contrarian effect.