In 2011, professor John Cochrane of the University of Chicago came up with the term “factor zoo”. A fan of factor investing, he felt, however, it had become “too much of a good thing”. Since about that time, the growth in new factor strategies purporting to deliver excess returns from systematic investing techniques has accelerated.
The proliferation of smart beta products and strategies was well underway back then. In Australia, this was nudged along by regulator bias to headline fee numbers, particularly by APRA, under interpretation of the MySuper legislation introduced in 2012. But prior to that, between 2010 and 2012, for instance, the academic literature identified the “discovery” of 59 new factors, according to researchers at FTSE Russell.
“By some counts, there are now over 300 factors in academic and practitioner literature. But in a zoo of 300, how can anyone sort through the herd and find a group of factors that will generate reliable, sustainable long-run premiums?”, FTSE Russell said in a client note in 2018.
The big index provider says it has three criteria to determine factors: they must be backed by solid academic research; they must have a clear economic rationale; and, they must be robust and unique. The firm’s assessment of factors includes three types of intuitive economic rationale:
- Rewarded risk. Certain factors have earned higher long-term returns as a reward for bearing greater risk
- Behavioural biases. Not all investors are perfectly rational all the time, generating opportunities for those who can take a contrarian view, and
- Structural impediments, such as market rules or restrictions that can make some investments off limits to certain investors, create opportunities for others who can invest.
Global funds manager Robeco, which refers to itself as “the investment engineers”, last month produced a short paper citing the academic literature, back to the famous work of Eugene Fama and Ken French in the early 1990s, titled: ‘How Many Factors Are There? Or, How to Navigate the Factor Zoo’.
The Robeco paper, by David Blitz, the head of quantitative research, and Matthias Hanauer, senior researcher, says that they share the concerns expressed in recent academic papers that many of the hundreds of factors that have been proposed over the past decades are either “redundant, lack robustness, or cannot even be replicated”.
The researchers say, however: “We do not find, however, that the entire factor zoo can be reduced to just a handful of factors. Although the small set of factors used in academic asset pricing models can serve as a very good starting point, that is not the end of the story.
“In our research, we find evidence of dozens of factors. These include factors that are wrongly dismissed or rejected, multiple factors to capture one broader phenomenon, factors based on non-standard data sources or with limited history, and ‘next generation’ factors, based on big data, machine learning or artificial intelligence.
“That said, for practical implementation purposes, it is common to categorize factors into a small number of strategic composite factors. The low-risk factor contains metrics such as volatility and beta, measured using different lookback periods and different data frequencies, but also distress risk indicators, such as distance-to-default and credit spreads. The value factor consists of all variables which measure price relative to fundamentals, such as book value, earnings and cash-flows. These ratios can be adjusted, for example, through distress risk or environmental footprint.
“The quality factor is essentially a mixed bag of company fundamentals, such as profitability, earnings quality and investment patterns. We agree with the academic perspective that these actually appear to be separate, distinct factors, but follow the industry convention to combine them under a common ‘quality’ header.
“For the momentum factor it is basically the other way around. Academics and smart beta index providers tend to see momentum as a single factor (price momentum), but in our research we find that it is better understood as an entire family of different sentiment-related factors, most notably price momentum and analyst revisions. Finally, there is another broad set of short-term factors. These are typically ignored by academics and index providers altogether, but we find them to be highly effective for trade timing purposes. This theme includes various reversal phenomena, signals based on short interest and signals derived from trading volume patterns.”
Starting from an academic zoo consisting of hundreds of alleged factors, they narrowed it down to several dozen that really work, and which can be organised into a small number of composite factors.