Cliff Asness: Low-Risk Investing without Industry Bets, we did a summary on this but for those who missed it…
Low-risk investing is based on the idea that safer stocks deliver higher risk-adjusted returns than do riskier stocks. This notion was first documented by Black, Jensen, and Scholes (1972), who found that the security market line was too flat relative to the capital asset pricing model (CAPM). For many, however, the intuition behind low-risk investing in stocks is captured by going long stodgy (but perhaps ultimately profitable) industries and by the related assumption that the returns are driven by value effects (e.g., Shah 2011).
Although there is nothing wrong per se with a factor that bets on industries, the tone of this criticism often conveys the idea that such bets, especially when passive (going in the same direction for long periods), are the result of path-dependent data mining or will somehow be particularly dangerous going forward. In any event, it is a common sentiment regarding these strategies and is meant to call into question their robustness and efficacy.
Discussion of findings. In our study, we explicitly tested how much of the benefit of low-risk investing comes from tilts toward or awayfrom industries versus stock tilts within an industry. We found that both types of low-risk investing work. Thus, contrary to conventional wisdom, we found that low-risk investing is not driven purely by low-risk industries—not even close—and is not driven by the value effect. Among all the low-risk strategies that we considered, those that take no industry bets are among the best.
There are many closely related forms of low-risk investing that focus on various measures: market beta (Black et al. 1972; Frazzini and Pedersen 2014), total volatility (e.g., Baker, Bradley, and Wurgler 2011), residual volatility (e.g., Falkenstein 1994; Ang, Hodrick, Xing, and Zhang 2006, 2009; Blitz and van Vliet 2007),2 the minimum-variance portfolio,3 and other related measures (for connections between these measures, see Clarke, de Silva, and Thorley 2013). In our study, we focused on market beta because it is the original measure and is most closely linked to economic theory.
In particular, we constructed betting-against beta (BAB) factors that invest long in a portfolio of low-beta stocks while short selling a portfolio of high-beta stocks (following Frazzini and Pedersen 2014). To make the BAB factors market neutral, the safe stocks on the long side of the portfolio are leveraged to a beta of 1 and, similarly, the short side of the portfolio is deleveraged to a beta of 1. Hence, the overall ex ante beta of a BAB factor is zero, and so its performance can be ascribed to the efficacy of low-risk investing, not to market movements.
The “regular” BAB factor in the literature is constructed by sorting stocks on their betas without regard to industries; thus, its performance could be driven by industry bets, stock selection within an industry, or a combination of the two. To determine which is more important, we constructed the following two new BAB factors-one with no industry bets and the other with only industry bets:
- Industry-neutral BAB. To see whether BAB works when the effects of industry tilts are eliminated, we constructed an industry-neutral BAB factor by going long and short stocks in a balanced way within each industry. We computed a BAB factor for each industry and diversified across these industries to produce an overall industry-neutral BAB factor. We tested a robust set of methods of diversifying across these essentially separate industry neutral BAB factors.
- BAB as a pure industry bet. To see how well low-risk investing does as a pure industry bet, we considered a BAB strategy that goes long and short industry portfolios. This extreme form of low-risk investing more closely fits the popular perception of a strategy that makes only big bets on industries.
