The wages of dishonesty: The supply of cheating under high-powered incentives

We use a novel design to identify how dishonesty changes through a broad reward range that, at the high end, exceeds participants' average daily wages. Using a sample of online Indian workers who earn bonuses based on six simultaneous coin flips, we show that the relationship between dishonesty and financial rewards depends on the incentive range. We find two novel effects as incentives exceed those used in most prior research. First, dishonesty increases and reaches its maximum as rewards increase from $0.50 to $3 per reported head and as earnings reach $15, indicating that rewards can indeed motivate more cheating when large enough. More importantly, we show that dishonesty declines at the highest reward levels (up to $5 per head) as individuals appear to engage in lower magnitudes of dishonesty. We detail how our results could be explained by a reference-dependent utility with internal costs of dishonesty that are convex in the magnitude of the lie, and show survey and simulation-based evidence that support this explanation.