Fama-French 5-factor model: why more is not always better

Three Robeco experts on empirical asset pricing give their views. They acknowledge the major contributions Fama and French have made to the literature in the past and so studied this new research with great interest. However the debate is set to continue – they take a critical view of this newly proposed model.

Nobel laureate Eugene Fama and Kenneth French have developed a 5-factor model¹ to describe stock returns by adding two new factors to their classic (1993) 3-factor model.² The 3-factor model consists of market risk, size and value. The size effect is that stocks with a small market cap earn higher returns than stocks with a large market cap, and was found to exist in the 1963-1990 period. The value effect is the superior performance of stocks with a low price to book compared with stocks with a high price to book.

Fama and French have now added profitability (stocks with a high operating profitability perform better) and an investment factor (stocks of companies with the high total asset growth have below average returns). Both new factors are concrete examples of what are popularly known as quality factors. The paper ‘A five-factor asset pricing model’ was published in the Journal of Financial Economics in April 2015. We asked three Robeco experts, who recently published a research article on this topic in the Journal of Portfolio Management, what they think of this new model. Pim van Vliet, Portfolio Manager Conservative Equites, David Blitz, Head of Quantitative Equity Research and Matthias Hanauer, Quantitative Researcher, discuss the implications.

Van Vliet sees the addition of two more quality factors as a big change from the old model. ”If you exclude market risk, the new model effectively doubles the number of factors to four. All these factors interact, which makes it more difficult to summarize the cross section of stock returns.”

David Blitz is also critical about the way the empirical research is approached. “This approach can even be considered as a form of tautology, because they use five factors to explain the returns of those same five factors.”

The two new factors (profitability and investment) are thus used to explain their own performance. I would prefer it if they showed that just a handful of factors can be used to explain the performance of the numerous factors found in the literature. They do this in the follow up paper, but it should form the basis for their study.

Van Vliet is more surprised by the factors they did not include. “The new model still ignores momentum, while this factor is widely accepted within academia and has been around for 20 years.”

They also omitted the low volatility factor, although for Blitz this is not such a big surprise. “There is a practical reason for excluding it, because it is not easy to combine with the market risk factor in the three factor model,” he says. “The market factor, which is similar to the beta factor of the capital asset pricing model, still assumes higher returns for higher risk, while a low-volatility factor would assume the opposite. An alternative approach would be to scrap the market factor altogether, but they did not choose this more radical step.”

Blitz argues that Fama and French have been too quick to add the two factors. “The two new factors they have added are relatively recent discoveries and the research of these factors in different markets and time periods is still limited.”

Hanauer says a different definition of two factors related to quality might be more appropriate. “They have added the two factors but it is unclear why they have chosen these precise definitions, “he says. “In my opinion, the definition of gross profitability given by Robert Novy-Marx³ would have been a natural choice.”

Adding two factors is a big step, but in some way it is only a modest step, thinks Blitz. “In the AQR paper, ‘quality minus junk’, twenty underlying quality variables are used, only two of which are selected by Fama and French. Why select two instead of twenty? Why select those two particular variables? A lot of questions remain unanswered.” Don’t get me wrong, he adds. “I am not against these two extra variables, but why have they given them a special status by putting them in a model, while a range of other variables are available?”

According to Hanauer, the two quality factors contradict earlier findings by Fama and French. “In their 2008 paper, Dissecting Anomalies,⁴ they stated that the asset growth and profitability anomalies are less robust. However, in their 5-factor model they use exactly this same asset growth variable in their investment factor.