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, Science|Business (paywall), Tech & Learning

• 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
We study speculative trading in Chinese and global commodity futures. Using a heterogeneous agent model, we decompose prices into contributions from fundamental traders, speculators, and manipulators, and estimate time-varying trader weights. Chinese markets have fewer fundamental traders and more speculators than their global counterparts. Among speculative activity, manipulators are more prominent in China. We find a gradual convergence between Chinese and global markets in the trader composition over time. Finally, we relate the increase in manipulators and the decrease in fundamental traders to a decrease in market efficiency.

History
Project Start in April 2025; First Version in December 2025

Abstract
We study how the financialization changed the informativeness of drivers for commodity futures returns. Based on a broad set of commodity futures, we show that the predictive power of fundamental, financial, and uncertainty variables for commodity returns substantially varies over the different stages of the commodity market financialization. We relate the changing informativeness of the drivers to the presence of financial hedgers, who diversify their portfolios with commodity investments. Using the highly financialized crude oil market as a laboratory, we find that when the market share of financial hedgers is high, the predictive power of financial variables for short-term commodity returns increases, while the informativeness of fundamental information for longer-term returns declines.

• 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)