I've been pointing out since the early 2010s that Thomas Piketty, Emmanuel Saez, and Gabriel Zucman have a bad habit of biasing their data construction to fit their politics.
Whenever they make a suspect call in how to handle stats, it ends up showing that inequality is on the rise. They almost never go in the other direction. The errors are non-random and almost invariably reinforce their political claims about rising income and wealth concentration since 1980.
Piketty did this throughout his book Capital in the 21st Century.
Piketty and Saez did this throughout their famous 2003 paper on inequality in the US.
Piketty, Saez, and Zucman Saez and Zucman did this in their Distributional National Accounts paper that Auten and Splinter dissected.
Saez and Zucman did this in their 2019 book on tax incidence and inequality.
Saez and Zucman did this in their flurry of "studies" and white papers to support Elizabeth Warren's wealth tax proposal, California Prop 40, a new wealth tax in France, and dozens of other similar examples.
A new AI report (linked below) audited 138 papers by Saez and Zucman. It found thousands of errors, 194 of which were serious. A whopping 97% of the serious errors went in the same direction: they showed inequality rising faster or made it higher.
But there's a bigger scandal at play here. Despite these clear biases, Piketty, Saez, and Zucman have enjoyed a stream of praise and prestige within the economics profession. They get expedited and friendly reviews in top econ journals even though a more serious peer review should have caught many of these errors (the Quarterly Journal of Economics is a recurring offender in this regard). Journals regularly give them a pass after the errors in their old work come to light - they're never expected to correct any of it, and just keep on using it as if nothing changed. Their non-peer reviewed data claims are also treated as gospel by major media outlets such as the New York Times and the Washington Post. And they get showered with prizes - Saez and Zucman are both Clark Medal winners from the AEA, and Zucman got his prize in 2023 after evidence of major data shenanigans in his work came to light in 2019.
Until now, those of us who have been trying to correct the record have had to wade through the PSZ trio's cluttered, disorganized, and often non-transparent replication packages (almost all of which are convoluted piles of Excel spreadsheets) to find the problems. I did a line-by-line replication of Piketty and Saez's entire 2003 US income inequality series to use one example, and found it was riddled with errors - almost every single one of which biased their results in favor of their political story.
We had the means to catch this stuff as it was happening, but the economics profession failed to do its job. And it failed precisely because its elite ranks wanted to believe the Piketty-Saez-Zucman story for political reasons.
Oct 28, 2026 · 8:44 PM UTC
Here's a link to a paper I wrote a decade ago documenting the biases and data problems in Piketty's book.
journal.apee.org/Private.Ent…
Here's where I do the same with Piketty's British data.
onlinelibrary.wiley.com/doi/…
Here's one where my coauthors and I replicated the first half of Piketty and Saez's famous 2003 inequality paper...and found it was riddled with errors and nonrandom biases.
onlinelibrary.wiley.com/doi/…
Here's where I document a slew of problems with all three, including Zucman's 2019 data malfeasance on US tax incidence.
independent.org/wp-content/u…
Here is my new paper with @VincentGeloso documenting the rest of the problems with Piketty-Saez from 1917-present.
papers.ssrn.com/sol3/papers.…
Here I am a few months ago on Saez and Zucman's activisting in support of the California Prop 40 billionaire tax.
wsj.com/opinion/the-deceptiv…
IIRC, afew weeks ago, you complained about the Byzantine structure of the PSZ data that made replication very difficult. Several people suggested that AI would make this easier, but you argued that it was not so straightforward.
Has this report changed your opinion? And has this report caught the same errors that you and other people pointed out before?
I've spot-checked this report and most of the issues it flags are the simpler kind that appear in descriptive text or a single page in their data spreadsheets.
I've had similar luck with getting AI to flag the simpler errors in various PSZ products. So I'd say it is partially successful in that respect.
Where it breaks down: the complex errors that require several layers to detect, and often require specialized knowledge.
I'll give one example that I caught in the famous PS 2003 income distribution paper. Due to changes in accounting practices and the tax base, PS have to apply a multiplier to pre-1940s data that estimates known missing filer income (caused by different married and single filer thresholds) generally hovering around the top 10% cutoff. How they do it causes some pretty substantial changes in all years prior, and the multiplier calculations are buried deep in multiple labyrinthine excel files. It turns out they made a pretty severe error in the base year they selected to calculate their multiplier, and it came from misreading IRS pdfs on how they sampled and reported a segment of filers in the year PS chose. Their methodological choice also increased their inequality measure in several years before the New Deal where a known issue with tax reporting accuracy from that era otherwise cut against this claim. I figured out the problem by comparing the 1939, 1940, and 1941 IRS records, retracing what the IRS had actually done and how it differed from PS, finding the correct data, and redoing their spreadsheets with it to calculate a new multiplier.
At least as of right now, AI would probably stumble somewhere in the excel spreadsheet cross-referencing with the IRS figures. And it would almost certainly fail at extracting the correct figures from comparing the 1939-41 IRS reports. So it probably wouldn't succeed at catching one of the more sophisticated and consequential problems of PS 2003.
A child could understand this stuff you ret@rd. If you don't think inequality is the biggest issue in the US today you are a ret@rd.
youtube.com/watch?v=QPKKQnij…
Hey @grok. Explain to this anonymous rando why his approach to conversation is likely to get him blocked. Be harsh, condescending, and uncharitable to him.
Please please publish the critiques/take downs of these papers. Or point links to where it has been done. My statistics professor used to say, if you torture the data long enough it will confess to anything. The single highest factor determining of a child will grow up in poverty is if there is ONE full time worker in the home. Lower income people are significantly more likely to be laid off or loose job in a recession. Prioritizing full employment and economic growth is the clearest and surest way to ensure upward mobility.
There might be problems with their papers, but this AI investigation is as biased and sloppy as the papers. Many of their so-called “major” errors seem to be driven by the political view of the person who wrote the prompts. This is a hit piece and I would want to read it carefully before promoting it.
Tyler Cowen posted about this study that used two frontier AI models to independently double check 138 economics papers. A third AI reviewed the results of the other two AIs. The vast majority contain significant or major errors.
Reading through the results, it's striking that the authors are focused on only one prime directive: Extracting money from the wealthy through taxation. One gets the sense that they believe that all wealth people a) under-report income, b) hide income, c) resist taxes, d) drive through loopholes to reduce taxes, e) are dishonest, f) aren't subjected to large enough tax rates. It's clearly a postulate in their world view that the wealthy are scoundrels who are unprincipled.
But if we apply the exact same ethics they ascribe to the wealthy to Zucman and Saez, we get a similar result in the integrity of their academic papers. That is, Zucman and Saez a) fail to be objective, b) omit disadvantageous facts, c) resist objective peer review, d) use foibles and emotions of sympathizers to get papers of inadequate substance published, e) are dishonest, f) are not held to a high enough standard of objectivity.
In the realms of finance and academic economics, sufficient auditing of the former is vastly easier and more likely than the latter. Hence it is far more likely that billionaires are adequately taxed (by objective standards) than it is that Saez and Zucman's works meet objective academic standards.
Along comes AI, without an amygdala or cerebellum to tell it want to do. Only a vast corpus of data and the iron clad rule to reduce inconsistencies between facts. Applying AI to economics papers is therefore highly likely to surface sneaky academic work, because it is less likely to have a normative objective to press.
