A0696
Title: Selection confidence sets for equally-weighted portfolios
Authors: Sandra Paterlini - University of Trento (Italy) [presenting]
Alessandro Fulci - University of Trento (Italy)
Davide Ferrari - University of Bolzano (Italy)
Abstract: Equally-weighted portfolios offer a straightforward investment rule that can deliver strong out-of-sample performance compared with optimized portfolios. Nevertheless, the uncertainty over which equally-weighted portfolio is truly optimal among all possible combinations is typically overlooked. The selection confidence set is introduced: The collection of equally-weighted portfolios that, at a chosen confidence level and under a specified loss function, are statistically indistinguishable from the unknown optimal portfolio. The selection confidence set challenges the conventional notion of a single best portfolio. Analogously to a traditional confidence set in parameter estimation, its cardinality reflects the degree of portfolio selection uncertainty: In the presence of noisy data, distinguishing among competing portfolios becomes difficult, resulting in a larger set. Using Monte Carlo experiments with different initial assumptions and empirical data sets, it is shown that the selection confidence set is a practical tool for investors who wish to assess the robustness of their allocation choices.
