When do "Nudges" Increase Welfare?

Policymakers are increasingly interested in non-standard policy instruments (NPIs), or "nudges," such as simplified information disclosure and warning labels. We characterize the welfare effects of NPIs using public finance sufficient statistic approaches, allowing for endogenous prices, market power, and optimal or suboptimal taxes. While many empirical evaluations have focused on whether NPIs increase ostensibly beneficial behaviors on average, we show that this can be a poor guide to welfare. Welfare also depends on whether the NPI reduces the variance of distortions from heterogenous biases and externalities, and the average effect becomes irrelevant with zero pass-through or optimal taxes. We apply our framework to randomized experiments evaluating automotive fuel economy labels and sugary drink health labels. In both experiments, the labels increase ostensibly beneficial behaviors but also may decrease welfare in our model, because they increase the variance of distortions