Research Review

The case for low-risk equity investing: evidence from 2011–2025
Raul Leote de Carvalho (BNP Paribas), et al.
July 2025
This paper investigates the performance of equity low-risk strategies since 2011, highlighting their ability to deliver strong risk-adjusted returns across diverse market conditions. We introduce a composite risk score that extends beyond volatility and demonstrate its effectiveness through empirical analysis. The study compares portfolio constructions, examines sector-level effects, and evaluates exposures to Fama-French factors. Results confirm the persistence of the low-risk anomaly and the presence of alpha unexplained by traditional risk premia, supporting the case for including low-risk strategies in long-term equity portfolios.

Forecasting Global Financial Crises through AI-Enhanced Sentiment and Economic Signals
Leopoldo Mazzilli (independent researcher)
August 2025
This paper proposes a novel framework for forecasting global financial crises by integrating AI-driven sentiment analysis with traditional macro-financial indicators. Leveraging large language models and machine learning algorithms, we extract public sentiment signals from unstructured digital sources-such as social media, financial news, and expert commentary-and combine them with structured variables like credit-to-GDP gaps, yield curves, and market volatility. We construct composite sentiment indices and feed them into an ensemble model that includes XGBoost and LSTM architectures, generating a dynamic Crisis Risk Score (CRS). The framework demonstrates strong predictive capabilities across major historical crises, including the 2007–2009 financial meltdown, the 2020 COVID-19 shock, and recent banking disruptions, outperforming conventional econometric models. Our results suggest that sentiment can act as a forward-looking indicator of systemic risk, especially when combined with economic fundamentals. This interdisciplinary approach contributes to a deeper understanding of financial instability and offers practical tools for central banks, regulators, and institutional investors seeking early-warning capabilities.

Strategic Style Allocation: Absolute or Relative?
Pim van Vliet (Robeco Quantitative Investments)
September 2025
This paper explores how investors can allocate strategically across equity styles, depending on their objective: absolute return or benchmark-relative performance. Defensive factors improve Sharpe ratios over full cycles but come with higher relative risk and weaker information ratios. By contrast, benchmark-relative strategies benefit most from return-oriented factors. Dynamic allocation rarely survives costs and requires unusually high skill. The most effective approach is integration: combining multiple factors and short-term signals within one framework reduces timing risk, lowers turnover, and improves both Sharpe and information ratios. These findings demonstrate how factor combinations can be tailored to meet different investment objectives, whether absolute or relative.

Predicting Extreme Returns with Fundamentals: A Machine Learning Approach
Richard Wang and Yi Liu (St. John Fisher University)
September 2025
This paper investigates whether accounting fundamentals and market-based variables can predict extreme stock returns — large gains (“rockets”) and severe losses (“torpedoes”) — over horizons of up to three years. Building on the two-stage contextual framework of Beneish et al. (2001), we apply eight machine learning algorithms to first distinguish extreme from normal firms and then separate rockets from torpedoes. Using 45 accounting and market-based predictors, the models achieve strong predictive performance, with XGBoost consistently outperforming others and Random Forest close behind. Out-of-sample tests on U.S. equities from 2013–2025 show that predicted rockets earn steadily rising abnormal returns, while torpedoes suffer persistent underperformance. The spread between the two groups exceeds 100% over 750 trading days and remains robust across probability thresholds, based on Fama-French five-factor adjusted cumulative abnormal returns. The study contributes to the literature by integrating modern machine learning algorithms into a two-stage design, employing a broad set of fundamental predictors, and extending prediction horizons of extreme returns to the long-run.

On the Macroeconomic Foundations of the Anomaly Zoo
Michael S. O’Doherty (University of Missouri at Columbia), et al.
August 2025
We apply modern asset pricing methods that mitigate omitted variable and measurement error biases to estimate risk premia for 190 candidate macroeconomic factors using a broad cross section of equity style portfolios. More than 40 macroeconomic factors carry statistically significant risk premia. Models that include tradable mimicking portfolios for these factors frequently outperform leading multifactor models in explaining CAPM anomalies. Our findings reveal a strong link between economic fluctuations and asset prices, with the empirically most impressive factors tied to NIPA aggregates and housing market activity.

Variable selection for minimum-variance portfolios
Guilherme V. Moura (Federal University of Santa Catarina), et al.
August 2025
Machine learning (ML) methods have been successfully employed in identifying variables that can predict the equity premium of individual stocks. In this paper, we investigate if ML can also be helpful in selecting variables relevant for optimal portfolio choice. To address this question, we parameterize minimum-variance portfolio weights as a function of a large pool of firm-level characteristics as well as their second-order and cross-product transformations, yielding a total of 4,610 predictors. We find that the gains from employing ML to select relevant predictors are substantial: minimum-variance portfolios achieve lower risk relative to sparse specifications commonly considered in the literature, especially when non-linear terms are added to the predictor space. Moreover, some of the selected predictors that help decreasing portfolio risk also increase returns, leading to minimum-variance portfolios with good performance in terms of Shape ratios in some situations. Our evidence suggests that ad-hoc sparsity can be detrimental to the performance of minimum-variance characteristics-based portfolios.