While mindful of the critics’ fears over the evolution of a ‘factor zoo’, Bernie Nelson, president of North America at Style Analytics in Boston, Massachusetts, says that a transparent and objective factor-based framework approach that acknowledges the academic background but doesn’t tie itself to age-old definitions allows for “fair and objective comparison of any equity funds regardless of investment approach or construction.”
Following on from his recent opinion piece for ETF Stream, in the first of a two-part interview Nelson explains how the company has pioneered the visualisation of a portfolio’s factor orientation using a transparent approach, "free from opaque and prescriptive multifactor regression-based models". But we start with the arguments around divergence for academic stricture and why this is a necessary element for today’s investment climate.
ETF Stream: You clearly have a differing view on the degree to which we need to adhere to academic definitions of factors - can you explain where you diverge in this respect?
While academic investment research is a valuable input for the investment community, a potential danger arises when academic models are proclaimed as the ultimate arbiters of equity investing. Academic studies on factors are based on empirical observations and not on repeatable lab experiments. Investing is as much art as science and there should be some wiggle room for different philosophies and approaches to understanding factors in equity portfolios providing they have sensible economic rationale; can work across equity markets; are applicable to most stocks; are not just explained by sector or country effects; and have practical investment application.
Investors are always trying to get an edge on other investors. Knowing what everyone else knows is not usually a recipe for outperformance, meaning that professional investors are constantly looking to dig deeper or dig elsewhere than others. No academic paper is going to be the final word on what constitutes a factor, nor a rigid permanent prescription for equity analysis or portfolio construction.
Our factor choice is based on a practical combination of academic research, our own backtesting for factor relevance, and the current practices and independent research of active equity managers. The factors we use are well known and well understood by equity investment professionals. We do include factors from key academic models and research papers but we also have other factors that investment practitioners still find to be highly relevant and which can be justified outside of the many-decade time spans of academic models. For example, some academics insist that book to price is the only factor that should be used to define value, while many investors still commonly use other value factors including earnings yield, sales to price, free cash flow yield, and EBITDA/EV. These value factors can perform quite differently from book to price and to each other over practical periods that portfolios are actually being assessed on, including multi-year horizons.
ETF Stream: Can you explain further some of the issues around book-to-price when it comes to the value factor? Why do you think this has become an issue?
Despite the academic insistence of the use of book-to-price, there is current debate as to continued relevance of (it) since stocks with high book-to-price have been performing very poorly for at least the past decade compared with other popular value measures such as high sales-to-price or high free-cash-flow yield. Academics may make sensible claims that this is not a long enough time frame to conclude that book-to-price is no longer relevant as a value factor. They could be right. We still include this factor. But good arguments have also been made by investment practitioners that book-to-price has become less relevant because of the rise of intangible assets, the increasing importance of brand value, share buybacks, and also changes in accounting rules such as the treatment of R&D as an operating expense.
Different value-based investment approaches may incorporate various value factors for different reasons, each of which is plausible and each with its own biases. We know that the investment processes of some quantitative value managers omit book-to-price entirely in their models, while some may use measures such as EBITDA to Enterprise Value exclusively to define value; and many will use a composite of value factors to define an overall value factor. Designing a factor framework that can compare the resulting funds from these approaches on a fair basis requires consideration of these different viewpoints and will inevitably involve some subjective judgment.
Even the set of factor categories can extend in investment practice beyond those suggested by academic models. Yield is one example. During the 1990s and through the dotcom period, the correlations between dividend yield and other value factors were strongly positive but began to fall in the mid-2000s, even dipping into negative territory post-financial crisis. More recently, although still positively correlated with value factors, dividend yield has exhibited more idiosyncratic behaviour during the very low-interest rate period across global economies and has decoupled somewhat from the main value factors.
In addition, some investment practitioners and quants use shareholder yield as a stock selection factor. This measure includes dividend yield along with net buyback yield and net debt paydown yield. So overall, it can be practical and relevant to reclassify dividend yield into a separate ‘yield’ category along with the other components of shareholder yield. Some investors may wish to simply consider these yield factors as other variants of value, others could legitimately claim there is a distinction and may value having a separate category in order to evaluate equity portfolios, while others may ignore yield entirely. All of these views are valid from our perspective.
ETF Stream: Do you think that academic adherence is limiting for the wider implementation of factors?
Some academics criticise the use of factors outside the handful of factors that have withstood severe statistical scrutiny over multiple decades. The claim is that these other factors are not proven persistent drivers of returns in the future. We would argue that the choice of factors for an effective portfolio analysis factor framework needs to incorporate contemporary investment practitioner thinking and practice. Investors may want to confirm the forward-looking structure of a portfolio and current factor intention without simply relying on estimated factor beta exposures based on backward-looking analysis calibrated over multiple decades. Investors may also wish to tune into the dynamics of changeable financial markets over practical investment horizons, rather than assume that history will repeat itself based on a small set of seemingly immutable long-term factors.
There are many models and approaches followed by different equity investment practitioners. While the continuation of historically observed factors may be argued through risk or behavioural based arguments that are expected to persist, there also has to be some humility that things can change or just that medium-term cycles may be affecting factor returns differently and that many investors care about those time periods as part of their realistic time horizon.
Designing a factor framework that can compare funds on a fair basis requires consideration of these different viewpoints and will inevitably involve some subjective judgment which is where actual investment management experience is invaluable, as is having a large practitioner client base across investment managers, asset owners, and consultants – many of whom conduct their own high quality and often acclaimed research. Academic research papers on factors should not, therefore, be the final word on what constitutes a factor, nor a prescription for equity analysis or portfolio construction.
Portfolio insights also need to be transparent, and easy to communicate and understand, in order to maintain confidence and trust when making investment decisions. For example, while we do incorporate regression techniques when researching factors for inclusion, our portfolio analysis framework for assessing factor orientation is not itself a backward-looking regression model.
We do empathize with the aversion to having a ‘factor zoo’ of hundreds of factors to contend with although we also don’t think the other end of the spectrum with less than a handful of factors is practical or reflective of current investment practice either. Our factor list is still relatively parsimonious. For example, our standard Style Skyline shows 23 distinct factors across seven factor categories. Not exactly a factor zoo. But we also show all of these factors separately such as book to price, earnings yield, and cash flow yield. These are not simply combined into a composite factor and neither are they part of a multi-factor risk model.
The second part of the interview with Bernie Nelson will appear on ETF Stream later this week where is discusses why the demise of the Capital Asset Pricing Model shows why there is a flaw in the arguments about sticking to the prescribed academic path.