Of myriad possible factors that academic researchers have published over the past several decades, only a handful are suitable for long-term investing by those of us who have a limited amount of time to monitor our portfolios and a reluctance to make frequent changes that incur fees. Therefore, we need a small handful of useful criteria with which to winnow down the possible options. For this, I will draw heavily on the work of Andrew Berkin and Larry Swedroe (2016) in their fabulous little book Your Complete Guide to Factor-based Investing. Berkin and Swedroe review the literature on which factors really matter and then sort them based on the five basic criteria. We will consider each criterion briefly below, summarizing what Berkin and Swedroe say about them.
According to the Efficient Markets Hypothesis (EMH), and price advantage that an investor can find in the market will quickly be arbitraged away by other investors. When everyone is trying to gain an advantage in the same way, the conditions that allow for the advantage disappear. Sometimes, not all market participants will notice the existence of the factor for a protracted period. If an investor can keep the idea secret, she may be able to capitalize on it long enough to become fabulously wealthy. The quants of the 1980s did this to spectacular effect. As bad as I want to do what they did, it can no longer be done with anything approaching the degree of profitability that existed then. The referees found out what was going on, and the playing field has been mostly leveled.
Some factors do manage to persist over time, but the list is small. Part of the reason that these manage to persist is that they take a long time to profit from. The tilts (another common word for factor) that remain always come with extra risk, and there is opportunity cost risk that keeps many investors away. The market is composed of many different types of investors, and many of them have a strong desire to profit quickly. The speed of compounding of initial capital is directly related to the return on the investment.
Day traders, almost by definition, need big moves to happen quickly. Swing traders are in the business of predicting big moves and nimbly entering and exiting investments at the correct time. Hedge fund managers and active managers are under immense pressure to perform on an almost daily basis, and they take on large amounts of arguably unnecessary risk to keep shareholders and clients happy. Classical economics defines utility in terms of the investor’s long-term best interests, but reality suggests that many of us act in accordance with our short-term desires and needs. This can spawn market inefficiencies that can last for a very long time, and that resurface in cycles.
Regardless of the cause of the market inefficiencies, it is possible for them to persist across markets, across time, and across economic conditions. As long as investors perceive risk in a factor, that factor can remain persistent. Most factors will go away or be reduced significantly, but some few remain.
A pervasive factor is one that holds true (provides predictive validity) across barriers such as markets, nations, economic conditions, asset classes, and even sectors. To be useful as an investment tool, a factor needs to be common enough for us to take advantage of it. If we must constantly be on the search for needles in the proverbial haystack, then the factor fails our basic test.
If a factor is robust, then it is easy to measure despite the use of various, differing operational definitions. In other words, no matter how we measure it, it remains relevant to our investment decisions. If we examine the construct (factor) value, we can use price to book, free cash flow, earnings, or sales and still get a meaningful measurement of what is going on with a company. Meaningful prediction requires consistent measurement, and the concept of robustness can be thought of as a factor’s resistance to measurement error.
Academic papers may suggest many factors that aren’t very useful to the retail investor because barriers prevent us from investing with a view to capitalize on it. For reasons that we’ve already discussed, massive gains in excess of beta are not likely to happen. When we use factors to inform our investment decisions, we are thinking in basis points, not double-digit percentages. Any factor that we utilize must beat buying an S&P 500 fund with very low costs. This means that we must consider any costs associated with assets suggested by our factor. If the factor necessitates spending a lot of extra money in fees, commissions, and other costs that detract from our bottom line, we must reject it.
Some investments are simply not available to the retail investor. There is no good method of investing in great works of art by old world masters, just as there aren’t many good ways to invest in timberland (value investors will, however, want to inspect the assets held by paper companies; the larger concerns own vast tracts of land and manage forests that span entire counties). Markets for rare objects and works of art are extremely illiquid, and the cost of entry is staggering. The Yale Endowment can take advantage of these massive and rare opportunities, be we as humble retail investors cannot.
I would prefer that Berkin and Swedroe had used a different term for this aspect. When they use it, they are referring to a logical, theoretical reason that the factor should pay a premium and that the premium should continue to exist in the future. My issue is that for many people, the idea of intuition has more to do with feelings than reasoning. If you read their text (which I highly encourage you to do), you will find that, most often, this amounts to ferreting out the risk that gives rise to the premium.
As computing power becomes ubiquitous and data mining more accessible, we must resist the urge to invest based on the latest models that may or may not have substance. Correlations are tricky things, and we must keep in mind that a mere correlation does not allow us to infer causation. Correlations can arise from random permutations in the data, and models that examine very short periods may be especially sensitive to random noise that looks like a pattern.
If we apply these criteria to the available “zoo” of possible factors that we might consider, we indeed have a short list. Berkin and Swedroe winnow it down to eight factors that match the criteria. Using similar rubrics, the MSCI research (2013) team narrowed it down to a mere six.
Last Updated: 6/25/2018