How does dishonesty diffuse? An experiment comparing spread and concentrated incentives

Abstract We experimentally test if the propensity to cheat changes when subjects face different distributions of the same monetary incentive. Such a manipulation represents a costless intervention in the many situations where multiple reports are required simultaneously, like timesheets or expense reimbursements. In our experiment, subjects have to roll a die in private five times and report the outcomes. In the LAST treatment, only the fifth reported roll determines the payoff, whereas in the SUM treatment, all five reports contribute proportionally to the payoff. The possible reward from reporting remains the same, but the two treatments differ in the effort required to obtain it: inflating one report or many. We find that subjects report significantly higher numbers in LAST compared to the average report in SUM. Looking at the total of reports, it does not differ by treatment since subjects do not inflate the reported rolls when there is no incentive. Dishonesty differs as there are more opportunities, having the positive effect of reducing the total profit obtained from cheating, but the negative effect of spreading the lies across all the opportunities. The total amount of cheating made remains stable across conditions, but it costs less when the incentive is spread on multiple reports.