One of the fastest growing themes in investment management today is factor-based investing. Although some investors may have good reasons to focus on a single factor or a set of particular factors, the general consensus is that one should hold a portfolio which is well-diversified across various factors that have been shown to deliver significant premiums in the long run.
A multi-factor allocation offers investors exposure to various factors, avoiding large concentrations in any single strategy that expose their capital to the risk of short-run underperformance of any given style.
Factor mixing versus multi-factor integration
The key question is whether to mix stand-alone factor ‘sleeves’ or to construct a bottom-up integrated factor portfolio. The factor mix approach entails allocation to single-factor strategies that are constructed to provide maximum exposure to a chosen factor. This approach is transparent, convenient for performance attribution, and also allows for tactical factor allocation.
In most cases, however, this approach ignores the fact that a strategy which targets one specific factor may implicitly bring along undesired negative exposures to other factors. As a consequence, mixing single-factor sleeves can result in sub-optimal and unintended exposures at the overall portfolio level. The proponents of the integrated approach use this as their main argument, claiming that one can achieve better performance characteristics by combining factors optimally from the outset. They do so by selecting stocks with the highest integrated factor scores. Some other advantages of this approach are transaction cost netting and a moderate decrease in turnover.
Generic versus enhanced factors
Another important distinction should be made between ‘generic’ single-factor strategies based on a single factor model, and ‘enhanced’ single-factor strategies that are designed to eliminate unintended exposures to other factors. Generic single-factor factor strategies tend to be suboptimal compared with enhanced factor strategies. For instance, some generic value strategies do not provide much exposure to the value premium to begin with. Other generic value factor strategies do provide a high degree of value exposure, but are only able to do so by accepting significant negative exposures to other factors.
Enhanced single-factor strategies provide many advantages of the single-factor sleeves, while overcoming the pitfalls mentioned above. For instance, adding some momentum and quality exposure to value strategies helps to avoid the so called ‘value traps’ (stocks that are cheap for a reason) and adding value to momentum or quality helps to avoid the most overpriced stocks. As the integration of other factors is only partial, the main return driver remains the selected factor premium. These ‘enhanced’ factor strategies can be further mixed into multi-factor vehicles, but interestingly, this approach has not been a subject of thorough investigation by the literature that explores the ways in which factors can be combined into multi-factor portfolios.
Which approach is better?
Investigating which approach is better - bottom-up integration or mixing single-factor strategies -Ghayur, Heaney, and Platt (2016) argue that a preference for one or the other ultimately depends on the main objectives of the asset owner. They find that both approaches are viable options.
Leippold and Rueegg (2017) consider a richer set of factor combinations, robust statistical tests, and longer time periods to conclude that integrated portfolios do not outperform the mixed ones. While they do find that the integrated approach lowers the overall portfolio risk through better portfolio diversification, this lower risk is accompanied by a lower return. In fact, any difference in risk between the two approaches can be explained by a higher exposure of the integrated portfolio to the low-risk anomaly. The authors find no evidence in support of one approach over the other.
To summarize, it seems that, empirically, there is not much difference between the integrated and the mixed approach.