Detecting Data Fraud: Miles and Miles of Fabrication

Does signing an honesty pledge at the top or bottom of a document increase honesty? That was the question researchers explored in a field study about reporting on a car’s mileage. But the data from that study may be fake.

A post from entitled, “Evidence of Fraud in an Influential Field Experiment About Dishonesty” breaks down the data and looks at several anomalies, including: an implausible distribution (with a histogram as evidence), the nature of numbers reported (whether it’s natural to report rounded numbers or not), differences in fonts, and several others.

The post is a fascinating look at forensic data analysis in which a careful examination of data can reveal potential irregularities. As the post concludes, open science helps, “Data should be posted.  The fabrication in this paper was discovered because the data were posted.”