Working Papers

Abstract
This paper investigates how the introduction of Large Language Models (LLMs) has affected academic writing in finance. Analyzing 41,489 articles from 34 finance journals, we find that readability declined and the use of LLM-associated terms increased following the release of ChatGPT at the end of 2022. These trends are more pronounced among authors from non-English-speaking countries and in lower-ranked journals. The effects are also stronger for authors affiliated with institutions in countries that place less emphasis on creativity, display lower skepticism toward technology, and have higher levels of moral ambiguity. Author-level analyses suggest that while LLM adoption increases publication quantity, it does not enhance publication quality or scholarly impact.

• Feature in Generative AI in Higher Education, CXO Advisory Group, ESSEC Knowledge, ESB

• Presentations (scheduled): Utrecht University (2025); Dutch Sustainable Finance Network Workshop (Rotterdam, 2025); Dutch Math Finance Afternoons (Utrecht, 2025); University of Gronningen (2025); Asia Conference on Business and Economic Studies (Ho Chi Minh City, 2025); AVSE FBNet (online, 2025); Radboud University Nijmegen (2025); Corporate Finance Days (Gronningen, 2025); a.s.r. (Zeist, 2025); German Finance Association Annual Meeting (Hagen, 2025); Vrije University Amsterdam (2025); Technische Universität Dresden (2025); Ruhr University Bochum (2025); University of Maastricht (2025); Audencia Business School (Nantes, 2026); emlyon (Lyon, 2026); KU Leuven (2026)

History
Project Start in February 2025; First Version in June 2025

Abstract
Financial Robo-Advisors (FRAs) enable households to participate in financial markets with a limited amount of money and without time or place constraints. While FRAs can help investors overcome behavioural biases, they also have disadvantages, such as relying on a limited number of inputs and lacking individualization. We conducted a systematic literature review on the nascent research on FRAs to synthesize previous research results. We identify two streams of literature: (1) asset management, which focuses on designing FRAs and improving the functioning of these machine advisors, and (2) behavioural finance, which investigates technology adoption and issues related to biased advice. Among other topics, future research should address why FRAs do not appeal to less financially literate people, who likely would benefit more than others from using FRAs.

Abstract
We investigate drivers of commodity futures returns using mixed-frequency vector autoregression. Slowing real economic activity and increasing macroeconomic uncertainty precede negative monthly returns. Stock markets predict daily but not long-term commodity returns. Information from these drivers yields significant trading profits, though predictability changes with financialization periods. In recent years, futures prices show less sensitivity to financial variables. Our findings indicate that financial hedgers (commodity index investors) enhance the predictive power of financial variables but reduce the informativeness of fundamental information for future commodity returns.

• Best PhD Paper Award, Annual Conference of the British Accounting and Finance Association (Doctoral Masterclass)

• Feature in Macrosynergy

• Keynote: V International Workshop, Higher School of Economics (Moscow, 2022, online); International Conference on Climate and Energy Finance (Weihai, 2024); 5th International Conference on Research in Management & Technovation (Hanoi, 2024)

• Presentations: ABN AMRO (2024); University of Basel (2024); German Finance Association Annual Meeting (Hohemheim, 2023); Annual Meeting of the Commodity & Energy Markets Association (Chicago, 2022); International Association for Applied Econometrics Annual Conference (London, 2022); 8th International Symposium of Environment and Energy Finance Issues (Paris, 2022); 5th Commodity Markets Winter Workshop (St. Johan im Pongau, 2022); 6th Vietnam Symposium in Banking and Finance (Hanoi, 2021, online); HvB PhD Seminar (Chemnitz, 2021); Utrecht University (2021); Annual Conference of the British Accounting and Finance Association (Doctoral Masterclass, online, 2021); Econometric Research in Finance (Warsaw, 2021); 10th INREC (Essen, 2021); TU Dresden (2018); TU Chemnitz (2018); Queen's University Belfast (2018)

History
Project Start in March 2018; First Version in October 2018

Abstract
We test the out-of-sample trading performance of model-free reinforcement learning (RL) agents and compare them with the performance of equally-weighted portfolios and traditional mean-variance (MV) optimization benchmarks. By dividing European and U.S. indices constituents into factor datasets, the RL-generated portfolios face different scenarios defined by these factor environments. The RL approach is empirically evaluated based on a selection of measures and probabilistic assessments. Training these models only on price data and features constructed from these prices, the performance of the RL approach yields better risk-adjusted returns as well as probabilistic Sharpe ratios compared to MV specifications. However, this performance varies across factor environments. RL models partially uncover the nonlinear structure of the stochastic discount factor. It is further demonstrated that RL models are successful at reducing left-tail risks in out-of-sample settings. These results indicate that these models are indeed useful in portfolio management applications.

• Presentations: European Financial Management Association (Cardiff, 2023); European Alternative Finance Research Conference (Utrecht, 2022); Cardiff FinTech Conference (2022); Finance and Innovation Symposium (Hanoi, 2022)