In the Olympic sport of modern pentathlon, athletes compete in five distinct events: running, swimming, equestrian, fencing and pistol shooting. Versatility is the name of the game, as only a well-rounded skill set will win the day. We feel that a similar approach should be taken with factor investing.
Any single factor is an expected driver of risk-adjusted outperformance over the long run, but during months or even years, individual factors will have their time to shine. Most well-known factors, like value or size, exhibit return patterns that can be quite different from each other. While some market participants try to time which factor(s) will outperform, we think most investors are best served by combining individual factors and taking advantage of their low inter-correlations. When one underperforms, the others are likely to help buoy a portfolio, similar to traditional equity diversification strategies.
The process by which factors are combined has a large impact on the final composition of a portfolio. One common multi-factor strategy, called a “composite” approach, is to sort a starting universe of stocks for individual factors into single factor indexes, then average the resulting indexes. For example, let’s take an investor who wants to buy U.S. equities with exposure to the value and size factors (a two-factor portfolio). They would sort the universe of U.S. equities for value stocks, creating a value portfolio, and then they would repeat the process to create a small-cap portfolio for exposure to the size factor. After creating the value and size portfolios, the investor would simply invest 50% in each.
However, unlike traditional equity diversification, simply adding together individually crafted factor indexes can produce sub-optimal results. The problem with the linear approach of a composite strategy is that it dilutes the overall factor exposure of a portfolio. Any given stock could score very well on a single factor, but poorly on others. Aggregating up to the portfolio level, this could leave investors with a portfolio that does not have much factor exposure at all.
We prefer multi-factor strategies that, like the pentathlon, require stocks to compete across five distinct factors and perform reasonably well in each “event.” We like the “tilt-tilt” approach, also called comprehensive factor, which tilts one factor onto another in succession. Taking the same theoretical investor who wants value and size exposure, the comprehensive approach scores U.S. equities individually on value (i.e., cheaper stocks receiving a higher score) and size (large market capitalization names receiving a lower score). Next, the scores are multiplied all together and used to adjust the relative weight of the stock. Companies that score well on all the different factors will see their weight in the portfolio increase, and companies that score poorly will have their weight reduced or even eliminated entirely.
The diluted factor exposure of a composite approach becomes clear when we look at the factor exposure of that approach compared to comprehensive factor approach. Using the Russell 1000 index as our base universe of U.S. equities, we compared three approaches: single factor indexes, the composite approach and tilt-tilt. Figure 1 shows the “excess” factor exposure of each approach—i.e., how much factor exposure each has that is higher than what could be found organically in the Russell 1000 index. As expected, the highest exposure to any single factor comes from the single factor targeted indexes, which sort stocks only for that one attribute and have negligible or even negative exposure to other factors.
But when it comes to the multi-factor approaches, only the comprehensive factor methodology maintains high exposure to each factor. The dilutive effect of combining single factor indexes means in the end the composite approach has negligible exposure to many factors.