Value by Design?

Abstract

Although academics and practitioners frequently refer in their work to equity value investing, no consensus exists as to what this style exactly encompasses. For a wide range of 3,168 alternative implementations (design choices) of what could all constitute value portfolios, the authors document the impact of parameter pertubations on risk-adjusted returns. The observed dispersion in Sharpe ratios allows the authors to identify the hierarchy of design choices and to better assess the degrees of freedom consumed within the strategy development process. The authors can therefore derive critical t-values that adjust for overfitting. This will prove to be useful in research governance and strategy selection. TOPICS:Style investing, statistical methods Key Findings • Design choices create large variations in risk-adjusted returns. Characterizing factor investing as (scientific or smart) betas with no alpha underestimates the importance of skillfull choices. • Among design choices, signal definition is most important in explaining risk–return differences. However, even inferior signals can lead to significant performance gains when combined with selected portfolio construction choices, indicating the potential for data mining. • Investment firms with credible research governance need more awareness of the many degrees of freedom their research teams consume daily. The statistical literature on multiple hypothesis testing can provide those tools.