Why do some nudges work and others not?

While nudges have recently gained popularity, many nudging interventions fail, and the effects of successful ones are often short-lived. We conjecture that the success of a nudge depends on how it interacts with the underlying economic incentives that determine the payoff-maximizing behavior of the decision-maker. For example, in the domain of tax compliance, a nudge is likely to be effective only if it is financially optimal for the taxpayer to pay the tax. To test our conjecture, we run a multi-period experiment in which we manipulate tax audit probability and nudge participants to report their income. In addition, we vary how often the nudge appears to test whether more frequent nudging increases long-run compliance. We observe that the first application of a nudge has a positive immediate effect on income reporting, irrespective of whether it is optimal to comply or not. However, subsequent nudges increase income reporting only if the nudge is aligned with the taxpayer's incentives. More frequent nudging in the direction opposite to incentives yields no effects on long-run compliance. Policy implications are discussed.