Constructing Long-Only Multi-Factor Strategies: Portfolio Blending versus Signal Blending

Abstract

We propose a framework for comparing and contrasting the portfolio blending and signal blending approaches for constructing long-only multi-factor strategies, based on exposure-matched portfolios. According to this framework, we expect the two approaches to perform similarly at low-to-moderate levels of factor exposures, due to a high degree of overlap between the two approaches. At high levels of factor exposures, due to better diversification, we expect signal blending to outperform portfolio blending. In empirical tests, which consider value, momentum, quality, and volatility factors across global equity markets, the expectation is validated for high levels of factor exposures. For low-to-moderate levels of factor exposures, however, we find that portfolio blending generates generally higher information ratios. This finding is largely driven by the interaction effects between the considered factors. Such interaction effects are overwhelmed by the high concentration and stock-specific risk in the portfolio blend at high levels of factor exposures. Furthermore, in the context of smart beta investing, portfolio blending may be better suited to achieve additional portfolio objectives, such as transparency in performance attribution, ability to time factors, ability to better control turnover and unrewarded risks at the individual factor level, and catering to certain asset owner governance considerations. However, the analysis and results reported in this article notwithstanding, it is also our view that an assessment of the relative merits of the two approaches needs to be conducted in the context of a manager’s specific investment process and key portfolio objectives pursued by an asset owner.