The authors introduce a five-factor asset pricing model that outperforms the well-known Fama–French three-factor asset pricing model in explaining stock returns. Surprisingly, when the two additional factors of profitability and investment are added to the original three-factor model, the value factor becomes superfluous. Although the five-factor model is not without its challenges, it is useful in describing the cross-sectional variance of the factors’ expected return.
Eugene Fama and Kenneth French have revised and expanded their original three-factor asset pricing model (Journal of Financial Economics 1993) to include two new factors: profitability and investment. They show that it performs better than their well-known three-factor model, although the revised five-factor model is not without its shortcomings. Small-capitalization stocks and their performance still pose a challenge to pricing models. Interestingly, the way the factors are defined does not influence the model’s performance.
How Is This Research Useful to Practitioners?
The authors’ original three-factor model was designed to improve on the well-known and much-used CAPM. The CAPM uses a beta factor (determined by the difference between the market portfolio’s return and the risk-free return) to help explain portfolio returns, so Fama and French added two more factors—size and value—that showed improved explanatory results. At the time of their 1993 article, these factors were two recognized patterns in average returns that were not explained by the CAPM (i.e., anomaly variables).
More recently, additional evidence from other researchers indicated that the original three-factor model was deficient because the model did not account for profitability and investment in relation to the variation in average returns. Motivated by this evidence, the authors added two additional factors to their original model to come up with a five-factor asset pricing model.
The authors estimate that the five-factor model explains between 71% and 94% of the cross-sectional variance of expected returns for the size, value, operating profit, and investment factors in the portfolios they examine. In applications for which the sole interest is abnormal returns, their investigation reveals that when the value factor is excluded from the five-factor model, the model with four factors performs as well as the five-factor model. Unpredictably, in the five-factor model the value component is redundant for describing average returns because the value return is captured by the exposure of value to other factors. But the five-factor model should be used if one is interested in using a value screen. Despite the five-factor model failing the Gibbons, Ross, and Shanken (Econometrica 1989) statistical test, it does perform well because the unexplained average returns for individual portfolios are nearly all close to zero.
How Did the Authors Conduct This Research?
Data suggest that stock returns are related to the book-to-market equity ratio as well as to profitability and investment. As a starting point, the authors use the dividend discount model to explain why these variables are related to average returns. With a bit of manipulation based on the dividend discount model, the authors are able to extract two additional factors, profitability and investment, to add to their three-factor model. They define profitability as operating profit minus interest expense divided by book equity, and they measure investment as the change in total assets divided by total assets.
The study covers 606 months of data from July 1963 to December 2013, which includes an additional 21 years of new data from when their original three-factor model was published in 1993. At the end of each June, stocks are allocated to various size groups using NYSE market-cap breakpoints. In addition, the other factors (i.e., value, operating profit, etc.) are segregated within their respective categories and ranked from low to high. The authors calculate the excess monthly returns of the factor portfolios over the one-month Treasury bill rate. Finally, they measure the standard deviations, t-statistics, correlations, regression intercepts, coefficients, and slopes of the portfolios they construct to analyze the data.
Although Fama and French have developed a new five-factor model and as popular as the Fama–French three-factor asset pricing model is, the question is whether the model will be as well received by investment practitioners and the financial community. Especially with the value factor falling away and being replaced by profitability and investment factors, it basically becomes a four-factor model. Somewhat surprising is that small-cap stocks still seem to be elusive to their model. I would be interested in knowing what other factors (in addition to momentum and liquidity) the authors may have analyzed before deciding on the five-factor model presented in their study.