A Study on Honesty Turned Out to Be Built on Fabricated Data. The Metadata Named the Author.
In 2012, a study on honesty was published in one of the most respected journals in American science. It claimed that having people sign an honesty pledge before filling out a form, rather than after, made them report more truthfully — a tidy, citable, real-world-tested finding. Nine years later, three outside researchers proved the underlying data was fabricated. The metadata even said who fabricated it.
Peer Review Doesn't Audit Your Spreadsheet
Academic publishing runs on a good-faith assumption that co-authors don't fabricate results and that peer review isn't built to forensically inspect raw data files — it's built to evaluate whether a claimed finding is plausible and well-argued, not to re-derive it row by row. The 2012 PNAS paper, co-authored by Shu, Mazar, Gino, Ariely, and Bazerman, described a field experiment with an actual auto insurer and became a staple example in behavioral economics — cited hundreds of times, taught in classrooms, and featured in Dan Ariely's own bestselling book on dishonesty.
Two Fonts in One Spreadsheet
In August 2021, the Data Colada team — Uri Simonsohn, Leif Nelson, and Joe Simmons — published an analysis of the study's underlying Excel file that left, in their words, no shadow of a doubt: duplicated rows appeared in two different fonts, Calibri and Cambria, a telltale sign of data pasted in after the fact rather than collected in the field. The file's own metadata showed who had created it and who had last modified it: Dan Ariely.
Nobody Was Checking, Because Nobody Checks
The paper had passed peer review, been cited for nearly a decade, and shaped a subfield's public understanding of how honesty pledges work — and none of that required anyone to look at the raw spreadsheet, because that has never been how the system worked. The fabrication surfaced only because Data Colada went looking for unrelated replication reasons, years after the paper's influence had already been spent.
An Entire Field Now Works Under Suspicion
PNAS retracted the paper in 2021. The fallout went well beyond one author: Data Colada's follow-on posts extended scrutiny to co-author Francesca Gino's broader body of work, Harvard Business School placed her on administrative leave, and the case became a central reference point in behavioral science's ongoing replication crisis. Every honest researcher in the field now works under heavier data-transparency requirements and a background presumption of doubt that one fabricated spreadsheet, sitting uninspected for nine years, did a great deal to create.
Filed from public reporting:
