Errors Found by AI in the Research of Emmanuel Saez and Gabriel Zucman (marginalrevolution.com)

In a groundbreaking audit, AI tools have scrutinized the research of economists Emmanuel Saez and Gabriel Zucman, revealing significant flaws across 138 papers. The audit, conducted independently by Claude Opus 5.5 and OpenAI's GPT-6 Astra, identified a staggering 2,035 problems, with 147 being serious enough to alter core findings. Notably, 69% of the papers contained at least one issue, and 97% of the major and moderate errors inflated the papers' claims. This alarming discovery raises critical questions about the reliability of the research that has underpinned key economic policy debates. The methodology employed in this audit showcases the transformative potential of AI in economic research, using advanced computational techniques to thoroughly assess the accuracy of headline numbers derived from the authors' data. The dual-audit approach not only highlights the need for rigorous verification processes but also underscores a broader trend where AI could reshape our understanding of past research and influence future studies. As the implications of these findings unfold, the economic community faces a crucial moment for re-evaluation, signaling both potential shifts in scholarly prestige and a reevaluation of the policies based on this work.