Style Research applies six criteria to identify a factor framework for effective fund identification, comparison, and research. Our “Effective Factors” blog series has been exploring these criteria further. Here we discuss the benefits of being selective about factors for effective fund analysis.
How are fund buyers supposed to know which factors will be important for comparing thousands of equity funds when the underlying investment approaches are so different? How can they filter funds that reflect their investment views?
Trying to distill which factors matter in equity markets has been a holy grail for finance researchers for decades. Each year academic and practitioner researchers publish new results of their latest factor backtests in top journals to a receptive investing audience hankering for any new edge they can get. Over the years, the factors emerging from these papers have accumulated well into the hundreds. In 2011, John Cochrane, former president of the American Finance Association, famously described this proliferation as a “factor zoo”.
Incorporating factors: Survival of the fittest?
As a provider of independent and objective factor-based portfolio analytics, we naturally keep an eye on this research, and conduct our own, to see if there is anything new that we should be considering for our own factor framework. Simply adding hundreds of factors to a fund assessment framework would create overly complex fund comparisons. They would also be hard to communicate and digest and would be a poor way to deliver a global standard for portfolio communication. Ironically, investors would likely scurry back to the over-simplistic style box, leaving fund sellers underserved with little to showcase their unique points of differentiation.
There has been a natural backlash against the factor zoo by savvy researchers. For example, the 2016 president of the American Finance Association, Campbell Harvey, cautioned strongly about many of the academic research results in his Presidential Address. He referenced a paper, “…and the Cross-Section of Expected Returns” (2014), that he co-authored with Yan Liu and Heqing Zhu, all from Duke University. They examined 316 factors collated from published articles in top journals and highly regarded working papers. Many of the results were criticized for data mining, and for committing arcane statistical sins such as such as “p-hacking”, selection biases, and failure to control for multiple testing. Applying a more rigorous approach that adjusted for these biases led to the conclusion that only a handful of these factors were statistically significant, with most vaporizing under this brighter light. The few significant survivors included factors in well-recognized style categories such as value, momentum, and volatility.
These findings certainly suggest that anyone involved in creating or assessing investment funds might want to be more cautious before loading up on factors that emerge from academic papers and journals. Researchers also don’t always agree on what the key factors or styles are. In addition, and perhaps more insidiously, they are still picking over essentially the same historic data. And since we now know that factor performance is episodic, knowing what will be important going forward will always be a challenge. So these meta-studies are not going to be the last word on factors.
Factors are not just for quants
Despite the recent negative sentiment towards active management, the investment industry is still heavily populated with skilled and experienced equity investment managers and equity analysts who are focused on individual company research. They may be less interested in the “p-hacking” biases of academic research papers but more interested in understanding, for example, the impact of Tesla’s financing of its negative free cash flows or how its clean energy business might impact operating margin. Their research leads to decisions to buy or sell individual stocks that still ultimately manifest themselves through a portfolio or fund. Those funds are more likely to sell to investors using stock stories rather than the justification of factor premia. But the most successful investors are known for having a consistent and disciplined approach. Even the most concentrated stock picking portfolios leave a style or factor footprint that describes the nature of the stock selection process. (See for example last year’s analysis of the Sequoia Fund for Citywire.)
An effective factor framework ought to reflect accurately the attributes and underlying details of these types of actively managed funds alongside the more quantitatively managed or passively managed. This is where multifactor risk models lose their footing. Their goal is to forecast portfolio risk. The factors that emerge from that methodology might do a decent job of risk estimation or delivering a portfolio’s predicted tracking error. But the opaque factors that result from the methodology are barely understandable by the investment analyst tasked with fundamental company research, or even by many quants who use different models. Certainly, the investing client doesn’t get it.
This leads to another realization: the species of factors selected from the factor zoo need to be clearly identified, and preferably with recognizable names. If the factor doesn’t retain the language of the investment practitioner then the fast assessment and effective communication to the client will suffer. Even well recognized, basic factors – such as book-to-price, ROE, beta, or 12-month momentum – still require individual testing and justification to ensure their relevance in any portfolio analysis.
Of course there are other considerations that are essential for factor selection. Factors should be measurable across most companies and not just be explained by sector or country effects. They should have an intuitive economic rationale, should be evident across different markets and should be investable or liquid. They should also not just be a byproduct of small or micro-cap performance as a result of an equally weighted approach. And the factors ought to work for long-only investors. Many academic papers assume long-short strategies which are much less common in practice. All of these criteria are fairly well known. But the factor selection also needs to be fair to all funds and not biased to one type of fund. This needs to be borne in mind when designing and selecting factors for a factor framework. The framework must be effective for both fund sellers and fund buyers to analyze and compare funds objectively and confidently.
More criteria for an effective factor framework
In our next and final post in our series exploring the Style Research criteria for an effective factor framework, we’ll discuss the requirement that a framework and its factors are easy to understand and communicate, ensuring that the approach fosters collaboration with stakeholders and clients.
Until then, to review the Style Research criteria for identifying an effective analytical factor framework, please check out our related infographic.