Risk Reduction in Style Rotation

Abstract

Dupleich Ulloa, Giamouridis, and Montagu investigate the potential improvement in the implementation of style rotation strategies by techniques addressing estimation errors. They select two approaches that have recently stood out in the statistics and econometrics literature and have been applied to portfolio construction. One approach builds on regularization methods, addressing estimation error by focusing on the weights of the constructed portfolios.The second method pools forecasts that are obtained across different observation windows, thus focusing on minimizing estimation error in the moments of the return distribution that may arise due to structural breaks. The authors conclude that overall benefits are derived by foregoing naive approaches, which in their dataset can be as significant as an improvement in the Information Ratio of about 54%; that is, improving from 0.65 (naive) to approximately 1 (dynamic).