Research

Working Papers

AI Regulation and Voluntary Public Disclosure: Early Evidence from the EU AI Act

Solo-authored job market paper

Abstract: "This paper examines whether and how AI regulation can influence firms’ voluntary public disclosure choice by examining AI-related disclosure responses to the proposed European Union Artificial Intelligence Act (EU AI Act). Although this regulation does not require any public disclosure about AI, it may affect disclosure through changes in business practices and increased salience of AI as a regulated activity. I find that European firms are, on average, less likely to disclose about AI after the proposal compared to U.S. firms. There is substantial variation in this effect: European firms are more likely to start disclosing about AI, but also more likely to stop such disclosure after the proposal compared to U.S. firms. Descriptives suggest that these disclosure choices are associated with underlying financing and innovation practices. Furthermore, content analysis using topic modeling shows a lower emphasis on risk by European firms compared to U.S. firms following the proposal. These findings hold across all levels of changes in business practices, while the level of salience only influences the choice to stop disclosing. These results suggest that regulation of AI can reshape the voluntary public disclosure choice, leading to a new disclosure equilibrium."

Narrative Disclosures of Loss Firms

with Jeroen Koenraadt and Edith Leung (supervisor)

Received Convergence funding from AI and data science in 2023 

Abstract: "Reported losses have become increasingly common among U.S. public firms. A loss can reflect weak operations, GAAP-mandated expensing of investments, or transitory charges, but income statement line items do not always reveal what drives a loss. We develop a scalable approach to examine whether the narratives accompanying reported losses provide context to distinguish among these economic states. Using topic modeling and cluster analysis of the Results of Operations sections of 10-K filings, we classify loss firms into distinct narrative profiles that differ in their accounting characteristics and future performance. Firms with narratives describing advanced-stage R\&D or operational losses outperform other loss firms, whereas narratives describing early-stage R\&D underperform for up to five years. These differences persist after adjusting performance for intangible investment and are not explained by established textual attributes. Our findings suggest that narrative disclosures help users interpret what a reported loss represents, beyond what line items reveal."

Mandatory Pay Disclosures in Job Postings

with Jeroen Koenraadt, Nico Lehmann and Edith Leung (supervisor) 

Link to the latest draft: papers.ssrn.com/sol3/papers.cfm?abstract_id=6431939 

Environmental Regulatory Fragmentation

with James P. Naughton and Clare Wang

Work in progress

Regulatory Fragmentation and Corporate Misconduct

with James P. Naughton and Clare Wang