Factor Momentum Is Real, Researchers Argue

It may be possible to predict which factor investing strategies will outperform in the future, recent research from AQR Capital Management suggests. 

Investment factors like value and size exhibit performance persistence over time — or the factor known as momentum. Factor portfolios that have recently outperformed tend to do well in the future, and those that have underperformed are likely to continue to suffer, according to a new paper from AQR researchers Tarun Gupta and Bryan Kelly.

The study is part of a growing body of evidence that supports the existence of factor momentum, as does a paper published last January by three authors from smart beta shop Research Affiliates.

Factor momentum is a “very pervasive” phenomenon observed among all kinds of factor-based portfolios, said Kelly, an AQR vice president and finance professor at the Yale School of Management. His and Gupta’s study, which focused only on equities, examined 65 characteristic-based factors including valuation ratios and size, investment and profitability metrics, among others.

“What we noticed after looking at all kinds of portfolios is that you see persistence in return,” Kelly told Institutional Investor. “All these factors and style portfolios exhibit their own time-series momentum.”

Such findings are useful for asset managers — particularly factor-investing firms like AQR — because they offer insight into which factors are most likely to outperform in the future. “It might be possible to time your investment strategies based on recent performance,” Kelly said. “It tells us you can potentially do better by using a dynamic strategy that puts slightly more weight on strategies that have recently outperformed.”

In other words, investors could possibly use momentum to time factors. Factor-timing has been an area of research and discussion at AQR for the last few years, with co-founder Cliff Asness publishing several papers and blog posts on the topic, including a 2017 post, “Factor Timing is Hard.” In general, Asness has cautioned against attempts to time factors based on their relative valuations. But the new research from Kelly and Gupta is different, because it focuses on the momentum of factor strategies instead of their valuations. 

[II Deep Dive: Cliff Asness Blasts Rob Arnott on Factor Timing]

“One fact that continues to be true is we do not find any reliable predictors of future factor performance based on relative cheapness or expensiveness,” Kelly noted.

In the new study, Kelly and Gupta tested the factor momentum strategy by timing individual factors based on past one-month returns. Momentum-based factor timing resulted in positive excess returns for 61 of the 65 factors, and statistically significant alpha for 47 of them. When all the factors were combined, the factor-timing strategy earned an annual Sharpe ratio of 0.84 — a result that only decreased slightly when the window of past returns was lengthened from one month to one year and five years.

Even after accounting for transaction costs, which Kelly and Gupta acknowledged would “eat into” the performance of any momentum strategy, they found that the factor timing strategy delivered outperformance. The finding is in contrast with recent research by Dimensional Fund Advisors, which suggested that fees and trading costs wiped out any benefits earned using momentum.

“My research shows an implementable benefit of using momentum,” Kelly said. “That’s true for cross-section momentum” — a version of the momentum factor commonly applied to stocks — “and factor time-series momentum.”

Still, although the study supports the existence of factor momentum, Kelly cautioned that more research and data were needed to prove that a factor-momentum-based strategy would work in practice. 

“It’s important to keep in mind that this research is being done in an exploratory sense,” he said. “In the end there’s still more work to be done.”