Cheating and Incentives: Learning from a Policy Experiment

We use a database generated by a policy intervention that incentivized learning as measured by standardized exams to investigate empirically the relationship between cheating by students and cash incentives to students and teachers. We adapt methods from the education measurement literature to calculate the extent of cheating, and show that cheating is more prevalent under treatments that provide monetary incentives to students (versus no incentives, or incentives only to teachers), both in the sense of a larger number of cheating students per classroom and in the sense of more cheating relations per classroom. We also provide evidence of learning to cheat, with both the number of cheating students per classroom and the average number of cheating relations increasing over the years under treatments that provide monetary incentives to students.