Factor investing: new evidence on long-term equity risk factors
Robeco has released a lengthy academic study that examines the significance of various equity factor premiums. The new research analyses 24 equity risk factors in a uniform and robust fashion, across multiple markets and over a 200-year timespan. The paper identifies those risk factor premiums that are statistically significant and those which might be subject to statistical p-hacking (due to bad sample sizes or other causes).
We've split this list of papers into two parts - the first segment discusses the foundational theory underlying equity risk factor premiums and the evidence for their existence, while the second batch covers recently released papers on factor premium-related topics, such as factor valuation, factor momentum, and multi-factor approaches.
EVIDENCE FOR FACTOR PREMIUMS
To be considered relevant, a factor must first and foremost be backed by ample empirical evidence. In the absence of such evidence, academic research on multi-asset factor premiums could suffer from ‘p-hacking’ (or ‘data mining’). Recent research by Robeco uses new and previously unused deep historical financial data. The results allay any p-hacking concerns.
(This is the paper referred to by the above blog.) Authored by experts from Robeco's quantitative equity team, this paper examines a host of factor premiums across 200 years of data to determine to what extent p-hacking has taken place for factor significance.
The report explains the academic underpinning to factor investing, and describes how a consensus has built around the belief that 4-6 key equity risk factors can generate long-term excess returns. We provide links to pertinent academic research and professional papers, covering each of the key equity style factors, and detail the different factors being deployed by eight leading investment managers and smart beta ETF index providers.
This paper compares several factor models against each other in order to rank asset pricing models by their max squared Sharpe ratio. It includes an examination of Fama and French's three-factor and five-factor models, as well as one that incorporates a momentum factor.
This is Fama and French's introduction of their five-factor model. Interestingly, this model's performance isn't sensitive to how these five factors are defined.
The authors point to potential problems with the five-factor asset pricing model, including how it doesn't pay heed to momentum or low volatility, and how competing models have already been proposed.
A new method is suggested to select factors that supposedly explain a cross-section of expected returns. This study also finds that the market factor is the most important factor when looking at expected returns of individual stocks.
In this paper, the authors look at quantifying backtest overfitting within alternative beta strategies. They look at the live results of some of these strategies and investigate the robustness of certain factor exposures.
GIven the 'zoo' of factors in existence, a methodology is suggested to identify robust, investable risk factors that are supported by research on an appropriate number of markets or time-frames and are not affected by small changes in the definition of the factor itself.
RECENT PAPERS AND RESEARCH
Much has been written about factor evaluation, but how are factors valued? This FTSE Russell paper covers factor differences in factor valuation metrics.
Lazard Asset Management reviews growth, value, sentiment, quality, and risk factor returns across multiple equity markets for the month of March 2019.
AQR Capital Management looks at factor momentum, creating a portfolio that times factor exposures based upon recent performance. They find that this strategy can also be deployed as an overlay to traditional factor strategies.
This paper peers into how factor strategies such as momentum, value, and low volatility are used in Brazil.
Recessions, Expansions and Factor Performance: Not Much of a Factor-Timing Strategy (Axioma blog, 2019)
Axioma ran a quick study on factor performance before and after the onset of recessions, given data on every recessionary period for the past 35 years, finding that it may not make sense to time factor exposures alongside changes in the economic cycle.
This study proposes a multi-factor approach to outperforming an index by removing both 'lottery' stocks and 'safety' stocks.
The authors propose a method for evaluating the explanatory power of new factors and their contribution to asset pricing, thereby weeding through the factor zoo to determine which factors are most significant.
Long-only factor investing can be a valuable way for investors to achieve their long-term return targets. But realistic expectations about risk factors as well as patience through periods of underperformance are both necessary in order to achieve the best results.
Cliff points out that although the benefits of combining factors into a diversified portfolio are large and well understood, one obstacle for many is gaining intuition about results.
FTSE Russell shows that a bottom-up approach to multi-factor index construction that is based upon multiple tilting provides advantages to a top-down approach based upon selection and weighting.
For compliance reasons, this paper is only accessible in the United Kingdom
Many common factor indices exhibit unintended exposures to unrelated factors because of the simplistic way in which they were constructed. This paper discusses how to eliminate these unwanted risks in factor portfolios.