Factor investing uses a rules-based approach to isolate assets with certain characteristics that are expected to deliver superior risk-adjusted returns. Examples are stocks that are inexpensive relative to their fundamentals, stocks with strong recent performance, low-risk stocks, or high-quality stocks. Strategic allocation to such factors has been shown to provide diversification benefits and improve risk-adjusted returns over the more traditional portfolio management approaches that explicitly allocate to countries and/or sectors.
Interestingly, a recent working paper by Brierè and Szafarz (Factor-Based v. Industry-Based Allocation: The Contes, 2016) provides results of a contest between factor-based and sector-based investing. The authors conclude there is no clear winner between the two approaches in the long-only context. We challenge their conclusion and argue that results of the mentioned study are crucially dependent on the choice of factors that they consider. In fact, we show that an explicit allocation to the well-established factor premiums dominates allocation to sectors regardless of the optimization objective that is used.
Brierè and Szafarz (BS2016) consider the Size, Value, Momentum, and Quality factors and ten sectors, as classified by Kenneth French. While the authors show that factor investing is superior to sector investing within the long-short context, sector investing does as well, or even better than factor investing in the long-only context; i.e. when short-selling is not allowed. More specifically, the authors show that factor investing is the superior approach when evaluated using certain performance metrics, such as attaining the highest return, while sector investing beats factor investing when other performance metrics are used as the relevant evaluation criteria. For example, the paper claims that sector investing has the potential to offer greater downside protection than factor investing as strategies based on certain sector allocations are exposed to lower absolute risk levels than the best possible factor allocation.
Sector investing lacks a theoretical foundation
We oppose the idea that sector investing can add as much, or even more value than factor investing, even in the long-only context. First, there is no theoretical foundation for sector investing. While certain sectors have historically outperformed other sectors, there is no reason to expect this pattern will continue. Factor investing, on the other hand, is based on a vast amount of academic evidence that shows the existence of several factor premiums, provides reasons to expect these factors to continue to earn a premium in the future, and shows that factor-based strategies have added value in portfolios in practice.
Second, regardless of differences in the theoretical foundations, we argue that the empirical findings of BS2016 crucially depend on their selection of factors, and that conclusions turn in favor of factor investing if a different choice of factors is made. For instance, at Robeco, we believe in four key factor premiums that, next to Value, Momentum, and Quality, also includes the Low-Risk factor. The reason why BS2016 find that sectors have more potential to provide downside protection could well be because they do not include the Low-Risk factor in their selection. Consequently, the sector investing approach can be tilted to defensive sectors like utilities, but in their setting the factor investing approach is restrained from allocating to the defensive segment of the market.
Factor investing has superior results
We have ran our own horse-race between factor investing and sector investing, but this time also including a Low-Risk factor in the contest. We find that factor investing is superior to sector investing no matter what metric is used for performance evaluation. Moreover, the factor portfolios used in this analysis are based on very generic factor definitions. Using more sophisticated factor strategies is likely to give even better results, making the case in favor of factor investing even more compelling.