Expanding the Fama-French Factor Model with the Industry Beta
Anatoly B. Schmidt (NYU Tandon School of Engineering)
August 2023
Recently it was shown that the news-based stock pricing model (NBSPM) outperforms the momentum-enhanced five-factor Fama-French model (FF5M) for a representative list of holdings of the major US equity sector ETFs both in-sample (Schmidt 2023) and out-of-sample (Schmidt 2022). The leading term in NBSPM besides the market (CAPM) beta is the industry beta estimated using returns of the relevant equity sector ETFs or industry ETFs. In this work, the industry beta is added to FF5M. It is found that the resulting model (FF5MI) is generally more accurate in-sample than NBSPM in terms of the mean squared error but not necessarily in terms of the mean absolute error. However, FF5MI is always inferior to NBSPM out-of-sample. This implies that the industry beta has the major impact on stock prices while the FF5M factors may yield an over-fitted model.

The Cross-Section of Factor Returns
David Blitz (Robeco Quantitative Investments)
May 2023
We explore the cross-section of factor returns using a sample of 150+ equity factors. Most factors exhibit a positive premium and a negative market beta in the long run. Factor themes with a clear positive beta, in particular low leverage and size, have no alpha after controlling for this beta exposure. The remaining factors generate most of their raw return in bear markets, which also explains half of their decay in the predominantly bullish post-2004 period. Beta-adjusting factor returns yields alphas that are not only higher but also considerably more stable. We also revisit factor performance cyclicality, establish a low-beta effect at the level of factors, and confirm the existence of seasonal and momentum effects in the cross-section of factor returns. Altogether, our insights into factor behavior aid the development of more robust factor-based investment strategies.

Factor Investing Funds: Replicability of Academic Factors and After-Cost Performance
Martijn Cremers (University of Notre Dame), et al.
November 2022
Do factor investing funds successfully capture the premiums associated with academic factors? We explore this question using the growing number of factor investing funds that seek to capture those premiums. While, on average, such funds do not outperform, we find that the factor investing funds with the portfolios that most closely match their academic factors—determined using our novel, holding-based ‘active characteristic share’ measure—significantly outperform those that less closely match. Furthermore, adjusting for stock size, we conclude that the answer to our question is “yes” for closely matching factor investing funds, which net of costs duplicate the paper performance of the long side of academic factors.

Exploring the Factor Zoo With A Machine-Learning Portfolio
Halis Sak (Shenzhen University), et al.
November 2022
Over the years, top journals have published hundreds of characteristics to explain stock return, but many have lost significance. What fundamentally affects the time-varying significance of characteristics that survive? We combine machine-learning (ML) and portfolio analysis to uncover patterns in significant characteristics. From out-of-sample portfolio analysis, we back out important characteristics that ML models uncover. The ML portfolio’s exposure alternates between investor arbitrage constraint and firm financial constraint characteristics, the timing of which aligns with credit contraction and expansion states. We explain and show how the credit cycle affects different characteristics’ ability to explain cross-sectional stock return over time.

Residual Factor Prediction Via Time Series-based Machine Learning
Jianlin Chen (University of Southampton)
August 2023
Machine Learning algorithms have been widely used and proven effective in financial markets. In this paper, we introduced a Machine Learning model set trained on the residual factors from the Fama-French three-factor model (Fama and French, 1992) to find significant alpha factors. To include more information that the linear factor models did not have, we used time series-based Machine Learning models, like tree-based models and Shallow Neural Networks with time series features. We used the predicted residual factor to construct quintile portfolios and found that it provided a significant alpha return even if style factors were controlled. It helped us find the latent information in the residual term and prove the lack of time series information in the factor models.

Anomaly Predictability with the Mean-Variance Portfolio
Carlo A. Favero (Bocconi University), et al.
July 2023
According to no-arbitrage, risk-adjusted returns should be unpredictable. Using several prominent factor models and a large cross-section of anomalies, we find that past pricing errors predict future risk-adjusted anomaly returns. We show that past pricing errors can be interpreted as deviations of an anomaly price from the mean-variance efficient portfolio. Price deviations constitute a novel anomaly-specific predictor, endogenous to the given heuristic mean-variance portfolio, thus providing direct evidence for conditional misspecification. A zero-cost investment strategy using price deviations generates positive alphas. Our findings suggest that cross-sectional models should incorporate the information in prices to capture the time-series dynamics of returns.

Learn To Use R For Portfolio Analysis
Quantitative Investment Portfolio Analytics In R:
An Introduction To R For Modeling Portfolio Risk and Return
By James Picerno