In the world of factor investing, industry tilts are often an afterthought. Factor investment strategies systematically target stocks with characteristics that have historically been associated with better risk-adjusted performance. But they often end up with industry weightings that differ from the market's. This article summarizes a study Morningstar conducted to evaluate whether such industry tilts contribute to the success of value, momentum, and low-volatility factor strategies, or whether they are an uncompensated source of risk.
Even if industry tilts contribute to a factor strategy’s outperformance, normalizing factor measures to reflect differences across industries may still make sense. Valuations and accounting practices are more similar within an industry than across industries. For example, technology stocks almost always trade at higher valuations than those in the utilities industry and are generally less asset intensive. In light of these persistent differences, industry-relative factor signals may offer more information about stocks’ future expected returns than unadjusted signals. Substituting larger intra-industry bets for industry tilts may be prudent if intra-industry stock selection offers a more favorable trade-off between active return and active risk (higher information ratio) than industry selection.
Summary
The results of this study suggest that:
- Value and low-volatility investment strategies demonstrate persistent industry tilts, which do not significantly enhance their performance. Investors can reduce active risk without significantly hindering performance by constraining these factors' industry tilts.
- Momentum strategies perform best when industry weightings are left unconstrained. This is because momentum's industry weightings are dynamic, allowing it to effectively capture short-term persistence in industry performance leadership.
Research Design
To assess whether an industry-relative approach to factor investing is prudent, this study investigates whether tilting toward industries with stronger value, low-volatility, and momentum characteristics provides better performance. In that vein, we constructed value, momentum, and low-volatility factor strategies applied to both individual stocks and entire industries. Each factor strategy measures 50 years of monthly returns from December 1966 through November 2016, using data from the French Data Library. The stock-level factors ignore industry membership, so they can have industry tilts that may contribute to their performance. However, comparing the performance of the stock- and industry-level factor strategies helps illustrate the impact of those industry tilts.
The French Data Library sorts all U.S. stocks listed on the New York Stock Exchange, American Stock Exchange, and Nasdaq exchange into deciles at the end of June each year based on their book/price ratios from the prior year-end. We measured the stock-level value factor performance as the return difference between a market-cap-weighted portfolio of stocks in the cheapest five deciles and those in the most expensive five deciles.
We applied a similar approach to measure the industry-level value factor performance. Instead of ranking stocks, we ranked the 12 Standard Industry Classification industries at the end of June each year based on their book/price ratios from the prior year-end. The industry-level value factor performance is calculated as the return difference between a market-cap-weighted portfolio of the cheapest six industries and the most expensive six industries.
The stock- and industry-level momentum and low-volatility factors follow a similar approach. The momentum factor portfolios for both the stock and industry level were formed based on their prior 12-month returns, excluding the most recent one, and rebalanced monthly. This is consistent with the standard academic measure of momentum. The low-volatility factor portfolios were also refreshed monthly, based on each group's daily volatility during the preceding 60 trading days. While several low-volatility strategies use a longer measurement period, this is how the French Data Library sorts stocks into deciles by volatility, so we replicated it for the industry-level low-volatility factor portfolios. This short window can create higher turnover than a longer lookback period and may be a noisier predictor of future volatility.
.png)
In Part 2 of this article, we will look at the results from our study.
:quality(80)/cloudfront-us-east-1.images.arcpublishing.com/morningstar/5FNGF7SFGFDQVFDUMZJPITL2LM.png)
:quality(80)/cloudfront-us-east-1.images.arcpublishing.com/morningstar/EOGIPTUNFNBS3HYL7IIABFUB5Q.png)
