Clarifying Fama and French
Steven Thorley, CFA, and his coauthors Roger Clarke and Harindra de Silva, CFA, didn’t start out to write an article about the Fama–French three-factor model. To hear Thorley tell it, the genesis of the team’s recent article in the September/October 2014 issue of the Financial Analysts Journal, “The Not-So-Well-Known Three-and-One-Half-Factor Model,” was actually an attempt to research the application of the risk parity paradigm to individual stocks. I had a chance to speak with Thorley about his research.
“It became apparent,” Thorley says, “that many people have misunderstandings” about the classic Fama–French model. “What people envision isn’t consistent with what Fama and French actually introduced.”
Listen to the full interview with Thorley below. A lightly edited transcript follows the audio.
Pat Light: Welcome to the Financial Analysts Journal Author Interview series. I am Pat Light, assistant editor at CFA Institute. I’m here today with Steven Thorley, CFA, to discuss his recent FAJ article, “The Not-So-Well-Known Three-and-One-Half Factor Model,” published in the September/October 2014 issue with his coauthors, Roger Clarke and Harindra de Silva, CFA. So, Steven, in this interview series, we invite our authors to discuss their articles with us. We’ll get started here with this first question. What was the practical issue or issues that motivated your research?
Steven Thorley, CFA: This one started actually with an intent to do some research on a different topic. As it turns out, we were looking at strategies associated with applying the risk parity paradigm to individual securities, individual stocks. In the context of doing that, you need a risk model, a basis on which to do the analysis, and it became apparent that many people have misunderstandings about what we call the not-so-well-known three-factor risk model by Fama and French. They have misunderstandings about it. Often, what people envision isn’t consistent with what Fama and French actually introduced and certainly not consistent with the way risk models are used in practice today.
So, what was your approach to solving that issue?
We ended up deciding to write this paper on the risk model itself. We went back to basics. I am a professor here at the Marriott School at Brigham Young University. We talk about the difference between in standard portfolio theory what’s called the “capital market line” and the “security market line” (SML). The capital market line is simply the idea of leverage — that is, you can take a risky asset and lever it up or down. The security market line is from the traditional capital asset pricing model. And the security market line is an empirical assertion that the higher-beta stocks will have higher returns on average.
In fact, there’s a specific prediction about the slope of that line. We couched the research, which is primarily empirical in the context of this security market line, and as is fairly well-known nowadays, although perhaps there is not a full awareness of it among practitioners. The security market line is not only not at the right slope — it’s actually negatively sloped — but we are also using the largest possible dataset on US equity markets over the last half century. So, effectively, higher-beta stocks do not have higher realized returns. The best statement was probably that they’re about the same as low-beta stocks. In fact, if you take a point estimate, the slope of this SML is actually negative. So, we used that as the context in which to present this empirical fact that high-beta stocks don’t have higher realized returns and really never have. And then we talk about some of the implications of that.
Talk a little bit about those implications and how you think they are going to influence practice in the field.
Hopefully, first, people will conceptualize the right risk model. What I mean by that is that beta is an important characteristic of a stock, just like value and growth, size, maybe even other factors that are being introduced. I hope they will incorporate data as a factor that is separate and distinct from just the market return. That’s where we get this three-and-one-half factor model. The three factors are beta, value, and size, and the half factor is sort of a tongue-in-cheek way of referring to the market itself without a beta coefficient in front of it. That would be the primary implication in practice, that people have a better understanding of factor models and this classic risk model, in particular.
I think one of the other implications is to do better performance attribution. There is a case study in the paper about the performance of the various industrial sector spiders. One could, as a numerical example, think that the consumer staples sector had a very large positive alpha over the last decade — an alpha of over 300 bps, 3.31 to be exact. But in fact, the consumer staples sector is composed of a lot of low beta stocks. Once you take beta into account, the implied alpha from recent history is essentially zero, 33 bps, to be exact.
On the other hand, with things like the technology sector, one might infer a negative alpha, substantially negative alpha over a decade of –1.65%. But again, when you take into account the fact that the technology sector is composed of a lot of high-beta stocks, then that alpha is again almost zero.
Hopefully, it will also allow investors and portfolio managers to do a better job of performance attribution.
Is there anything else you’d like to add before we close out here?
Yes, maybe one thing. The research, because of the process of publication, is now a bit dated, about a year-and-a-half old. The calendar year 2013 was again a good illustration of these phenomena. Using a large-cap index, the US stock market went up just a little over 30% in the calendar year 2013. That would suggest, according to standard CAPM thinking, that a stock with a beta of, say, 1.3, would go up, say, 40%, whereas a stock with a beta of say 0.7 might only go up 20%. That wasn’t at all the case. In fact, there was, in the calendar year 2013, very little difference between the return on high-beta and low-beta stocks. The high-beta stocks did have a slight advantage of 2.4% but nothing like the gap that would be suggested by what I consider a misapplication of risk models and performance measurement. So, the “low-beta phenomena,” as it is now called, is alive and quite well in the current data.
That’s always fascinating, to hear that the research that someone does continues to be relevant a couple of years later and that we see the results play out in practice over the year during which we’re publishing an article. Steven, we really appreciate you taking the time to share your research and your ideas with us. We encourage all of you to read this article and articles from other CFA Institute publications on our website, cfapubs.org. You can also follow us on Twitter, @cfapubs. Thank you for listening, and Steven, we appreciate your time.
Thank you. Good to be here.
If you liked this post, don’t forget to subscribe to the Enterprising Investor.
All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.