A discrediting enterprise before a crucial vote in California? The story begins October 28 on Marginal Revolution, the economic blog of Tyler Cowen, professor at George Mason University. In a post entitled “Errors in the research of Emmanuel Saez and Gabriel Zucman”, he relays the results of an audit which was sent to him by an anonymous reader, as reported by L’Express.
In its first version, the “audit” claims to have examined 138 publications by Emmanuel Saez, Gabriel Zucman and their co-authors. It identifies 2,035 “problems”, including 147 likely to modify an important result. According to his calculations, 69% of publications would include at least one problem and 97% of major or moderate problems with an identifiable direction would lead to overestimating the researchers' conclusions.
But these figures have evolved. In a version updated on October 29, the audit now distinguishes 53 defects considered indisputable and 94 methodological choices that are questionable, but defensible. These results remain those of the audit and have not been confirmed by independent scientific expertise.
How was the “audit” carried out?
According to the site, two artificial intelligence models, Claude Opus 5.5 and GPT-6 Astra, examined the posts separately. They would have had access to the articles, their appendices and the available replication files to try to reproduce the researchers' calculations. A third AI would then reexamine the reported issues, attempt to refute them, and rank them according to their importance. However, the verifications are based only on accessible data. Confidential information to which only researchers have access could not be consulted.
The updated version also states that the models received “a corpus of 131 published reviews from other economists provided in advance, in order to test known objections rather than rediscover them. » The audit does not, however, constitute an independent expertise carried out according to the usual procedures of a scientific journal.
Who is behind this audit?
The identity of the author has not been revealed. We know neither his background nor his skills in economics or statistics. The work has also not been evaluated by independent researchers as part of a scientific publication.
Tyler Cowen is an American economist and professor at George Mason University. Holder of a doctorate from Harvard, he launched the Marginal Revolution blog in 2003, with Alex Tabarrok. He also hosts the Conversations with Tyler podcast.
He is also opposed to Proposition 40, an exceptional tax of 5% on the taxable net wealth of billionaires residing in California on January 1, 2026. Developed by four academics, including Emmanuel Saez, it must be put to the vote on November 3, 2026. In a post published on October 19, Tyler Cowen estimates that this measure could ultimately cost the state more than it would bring in, by pushing certain billionaires to leave California.
Critics, including from their opponents
Several economists have challenged the audit method, including researchers who disagree with the work of Saez and Zucman. L’Express notably cites Sylvain Catherine, professor of finance at the University of Pennsylvania (Wharton), and Antoine Lévy, economist at Berkeley. The latter describes a large part of the audit as “bullshit”.
Faced with this controversy, Emmanuel Saez and Gabriel Zucman reacted on
A controversy in the middle of the campaign against the tax on billionaires
Scientific disagreements over billionaire taxes are old. The work of Saez and Zucman was in fact already debated before this audit. Criticized in particular by Gerald Auten and David Splinter on their measurement of inequalities in the United States. A study published in 2025 in the Journal of Finance by Sylvain Catherine, Max Miller and Natasha Sarin also challenges their calculation method.
