Matthias Xaver Hanauer
Research Subjects
Matthias Xaver Hanauer has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- David Blitz
Is the chief researcher at Robeco, Quantitative Investments division, in Rotterdam, the Netherlands
5 shared publications David Blitz
Robeco Quantitative Investments
4 shared publications- Christoph Kaserer
Technical University Munich
3 shared publications - Clint Howard
Is a quant researcher at Robeco, Quantitative Investments division, in Rotterdam, Netherlands
2 shared publications - Laurens Swinkels
Erasmus University Rotterdam
2 shared publications - Marc Steffen Rapp
University of Marburg - School of Business & Economics; University of Marburg - Marburg Centre for Institutional Economics (MACIE)
2 shared publications
Latest Publications
- Open Access
- 10 March 2026
- Published by Elsevier in SSRN Electronic Journal
The Illusion of the Carbon Premium
Carbon that has not yet been emitted should not be used to predict stock returns. When emissions are not lagged, we closely replicate the positive carbon premium documented by Bolton and Kacperczyk (2021). Using only carbon emissions data available to investors at the time of portfolio formation, however, we find no positive relation between firms' absolute carbon emissions and stock returns. The reported carbon premium is therefore driven by a look-ahead bias, as estimated emissions are mechanically linked to future revenues when emissions data are not sufficiently lagged. We further show that regression-based estimates of the carbon premium are sensitive to skewed dependent variables and multicollinearity with standard controls. When greenness is measured by carbon intensity rather than absolute emissions, green firms have outperformed brown firms. We provide out-of-sample evidence for a positive green-minus-brown factor over the period 2020-2025, despite the strong performance of energy stocks during the first year of the Russia-Ukraine war. Unlike earlier periods, this recent outperformance is not driven by a further widening in the cost of capital of carbon-intensive firms.
- 10 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Illusion of the Carbon Premium
Carbon that has not yet been emitted should not be used to predict stock returns. When emissions are not lagged, we closely replicate the positive carbon premium documented by Bolton and Kacperczyk (2021). Using only carbon emissions data available to investors at the time of portfolio formation, however, we find no positive relation between firms' absolute carbon emissions and stock returns. The reported carbon premium is therefore driven by a look-ahead bias, as estimated emissions are mechanically linked to future revenues when emissions data are not sufficiently lagged. We further show that regression-based estimates of the carbon premium are sensitive to skewed dependent variables and multicollinearity with standard controls. When greenness is measured by carbon intensity rather than absolute emissions, green firms have outperformed brown firms. We provide out-of-sample evidence for a positive green-minus-brown factor over the period 2020-2025, despite the strong performance of energy stocks during the first year of the Russia-Ukraine war. Unlike earlier periods, this recent outperformance is not driven by a further widening in the cost of capital of carbon-intensive firms.
- Open Access
- 30 September 2025
- Published by Elsevier in Economics Letters
- Vol. 257
Model complexity and the performance of global versus regional models
- 30 September 2025
- Published by Elsevier in Economics Letters
- Vol. 257
Open AccessModel complexity and the performance of global versus regional models
- 15 February 2025
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 51 (6) , 28-50
Formula Investing
This article evaluates the effectiveness of four popular investing formulas—the F-Score, Magic Formula, Acquirer’s Multiple, and Conservative Formula—within a unified framework over an extensive period. Each formula generates significant raw and risk-adjusted returns, primarily by providing efficient exposure to well-established style factors. No single formula, however, consistently outperforms across all metrics. The Acquirer’s Multiple achieves the highest returns for top decile portfolios, the Conservative Formula leads in CAPM alpha and return spread, and the Magic Formula exhibits the highest remaining alpha after adjusting for common factors. Although all formulas remain successful for concentrated long-only portfolios in the post-2000 period, we observe some performance decay relative to earlier periods, underscoring the need for continuous innovation in investing strategies.
- 15 February 2025
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 51 (6) , 28-50
Formula Investing
This article evaluates the effectiveness of four popular investing formulas—the F-Score, Magic Formula, Acquirer’s Multiple, and Conservative Formula—within a unified framework over an extensive period. Each formula generates significant raw and risk-adjusted returns, primarily by providing efficient exposure to well-established style factors. No single formula, however, consistently outperforms across all metrics. The Acquirer’s Multiple achieves the highest returns for top decile portfolios, the Conservative Formula leads in CAPM alpha and return spread, and the Magic Formula exhibits the highest remaining alpha after adjusting for common factors. Although all formulas remain successful for concentrated long-only portfolios in the post-2000 period, we observe some performance decay relative to earlier periods, underscoring the need for continuous innovation in investing strategies.
- Open Access
- 1 January 2025
- Published by Elsevier in SSRN Electronic Journal
Do Machine Learning Models Need to Be Sector Experts?
We examine heterogeneous return predictability at the industry level using machine learning models trained on a comprehensive set of firm characteristics. We compare uniform (“Generalist”) models with industry-specific (“Specialist”) models and introduce a “Hybrid” model that combines Generalist and Specialist forecasts using a simple meta-learning framework. This Hybrid model yields significant improvements in out-of-sample return predictability and portfolio performance relative to both alternatives. Additional analyses using international data corroborate these findings. Overall, our results suggest that the Hybrid model benefits from a higher signal-to-noise ratio by balancing industry-level specialization and cross-sectional generalization.
- 1 January 2025
- Published by Elsevier in SSRN Electronic Journal
Open AccessDo Machine Learning Models Need to Be Sector Experts?
We examine heterogeneous return predictability at the industry level using machine learning models trained on a comprehensive set of firm characteristics. We compare uniform (“Generalist”) models with industry-specific (“Specialist”) models and introduce a “Hybrid” model that combines Generalist and Specialist forecasts using a simple meta-learning framework. This Hybrid model yields significant improvements in out-of-sample return predictability and portfolio performance relative to both alternatives. Additional analyses using international data corroborate these findings. Overall, our results suggest that the Hybrid model benefits from a higher signal-to-noise ratio by balancing industry-level specialization and cross-sectional generalization.
- Open Access
- 2 December 2024
- Published by Elsevier in SSRN Electronic Journal
Design choices, machine learning, and the cross-section of stock returns
We fit more than 1,000 machine learning models to predict stock returns, systematically varying key design choices across algorithms, target variables, feature sets, and training methodologies. The resulting nonstandard error of portfolio returns exceeds the standard error by 59%, with monthly mean top-minus-bottom portfolio returns ranging from 0.13% to 1.98%. These findings demonstrate that design choices are a first-order source of uncertainty in machine learning return predictions, and we provide recommendations for model design. We further identify the conditions under which non-linear models outperform linear models. Finally, we document that superior design choices persist over time and are identifiable in real time.
- 2 December 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessDesign choices, machine learning, and the cross-section of stock returns
We fit more than 1,000 machine learning models to predict stock returns, systematically varying key design choices across algorithms, target variables, feature sets, and training methodologies. The resulting nonstandard error of portfolio returns exceeds the standard error by 59%, with monthly mean top-minus-bottom portfolio returns ranging from 0.13% to 1.98%. These findings demonstrate that design choices are a first-order source of uncertainty in machine learning return predictions, and we provide recommendations for model design. We further identify the conditions under which non-linear models outperform linear models. Finally, we document that superior design choices persist over time and are identifiable in real time.
- Open Access
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Formula Investing
This study evaluates the effectiveness of four popular investing formulas—the F-Score, Magic Formula, Acquirer’s Multiple, and Conservative Formula—within a unified framework over an extensive period. Each formula generates significant raw and risk-adjusted returns, primarily by providing efficient exposure to well-established style factors. However, no single formula consistently outperforms across all metrics. The Acquirer’s Multiple achieves the highest returns for top decile portfolios, the Conservative Formula leads in CAPM alpha and return spread, and the Magic Formula exhibits the highest remaining alpha after adjusting for common factors. While all formulas remain successful for concentrated long-only portfolios in the post-2000 period, we observe some performance decay relative to earlier periods, underscoring the need for continuous innovation in investing strategies.
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessFormula Investing
This study evaluates the effectiveness of four popular investing formulas—the F-Score, Magic Formula, Acquirer’s Multiple, and Conservative Formula—within a unified framework over an extensive period. Each formula generates significant raw and risk-adjusted returns, primarily by providing efficient exposure to well-established style factors. However, no single formula consistently outperforms across all metrics. The Acquirer’s Multiple achieves the highest returns for top decile portfolios, the Conservative Formula leads in CAPM alpha and return spread, and the Magic Formula exhibits the highest remaining alpha after adjusting for common factors. While all formulas remain successful for concentrated long-only portfolios in the post-2000 period, we observe some performance decay relative to earlier periods, underscoring the need for continuous innovation in investing strategies.
- 24 November 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 50 (3) , 11-31
Factor Zoo (.zip)
The number of factors allegedly driving the cross-section of stock returns has grown steadily over time. We explore how much this “factor zoo” can be compressed, focusing on explaining the available alpha rather than the covariance matrix of factor returns. Our findings indicate that about 15 factors are enough to span the entire factor zoo. This evidence suggests that many factors are redundant but also that merely using a handful of factors, as in common asset pricing models, is insufficient. While the selected factor styles remain persistent, the specific style representatives vary over time, underscoring the importance of continuous factor innovation.
- 24 November 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 50 (3) , 11-31
Factor Zoo (.zip)
The number of factors allegedly driving the cross-section of stock returns has grown steadily over time. We explore how much this “factor zoo” can be compressed, focusing on explaining the available alpha rather than the covariance matrix of factor returns. Our findings indicate that about 15 factors are enough to span the entire factor zoo. This evidence suggests that many factors are redundant but also that merely using a handful of factors, as in common asset pricing models, is insufficient. While the selected factor styles remain persistent, the specific style representatives vary over time, underscoring the importance of continuous factor innovation.
- 14 September 2023
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 5 (4) , 40-65
The Term Structure of Machine Learning Alpha
Machine learning (ML) models for predicting stock returns are typically trained on one-month forward returns. Although these models show impressive full-sample gross alphas, their performance net of transaction costs post-2004 is close to zero. By training on longer prediction horizons and using efficient portfolio construction rules, the authors demonstrate that ML-based investment strategies can still yield significant positive net returns. Longer-horizon strategies select slower signals and load more on traditional asset pricing factors but still unlock unique alpha. The authors conclude that design choices are critical for the success of ML models in real-life applications.
- 14 September 2023
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 5 (4) , 40-65
The Term Structure of Machine Learning Alpha
Machine learning (ML) models for predicting stock returns are typically trained on one-month forward returns. Although these models show impressive full-sample gross alphas, their performance net of transaction costs post-2004 is close to zero. By training on longer prediction horizons and using efficient portfolio construction rules, the authors demonstrate that ML-based investment strategies can still yield significant positive net returns. Longer-horizon strategies select slower signals and load more on traditional asset pricing factors but still unlock unique alpha. The authors conclude that design choices are critical for the success of ML models in real-life applications.
- Open Access
- 12 June 2023
- Published by Elsevier in SSRN Electronic Journal
The Term Structure of Machine Learning Alpha
Machine learning (ML) models for predicting stock returns are typically trained on one-month forward returns. While these models show impressive full-sample gross alphas, their performance net of transaction costs post 2004 is close to zero. By training on longer prediction horizons and using efficient portfolio construction rules, we demonstrate that ML-based investment strategies can still yield significant positive net returns. Longer-horizon strategies select slower signals and load more on traditional asset pricing factors but still unlock unique alpha. We conclude that design choices are critical for the success of ML models in real-life applications.
- 12 June 2023
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Term Structure of Machine Learning Alpha
Machine learning (ML) models for predicting stock returns are typically trained on one-month forward returns. While these models show impressive full-sample gross alphas, their performance net of transaction costs post 2004 is close to zero. By training on longer prediction horizons and using efficient portfolio construction rules, we demonstrate that ML-based investment strategies can still yield significant positive net returns. Longer-horizon strategies select slower signals and load more on traditional asset pricing factors but still unlock unique alpha. We conclude that design choices are critical for the success of ML models in real-life applications.
- Open Access
- 29 May 2023
- Published by Elsevier in Journal of Financial Markets
- Vol. 65
Surprise in short interest
- 29 May 2023
- Published by Elsevier in Journal of Financial Markets
- Vol. 65
Open AccessSurprise in short interest