To see whether this signature effect could be harnessed to reduce cheating, Shu’s team enrolled 101 college students and employees in performing two self-reported tasks: solving math problems correctly in exchange for money, and claiming reimbursements for expenses on a library trip.
For each task, test participants filled out a claims form. Some signed at the bottom, others at the top, and others didn’t sign at all. Top-signers reported solving fewer problems, and claimed fewer expenses, than the other groups.
Considering that we humans lie kind of a lot, it's not unreasonable to think that there's a pretty serious truth gap on reporting mileage, reimbursements, and other things. If you're handing a form like this to somebody, consider having them sign at the top rather than at the bottom.
And hey, maybe Harvard should try that in its government classes.