WP-022: Hunt Allcott and Judd B. Kessler, "The Welfare Effects of Nudges: A Case Study of Energy Use Social Comparisons" (Revised October 2016)

“Nudge”-style interventions are typically evaluated on the basis of their effects on behavior, not social welfare. We use a field experiment to measure the welfare effects of one especially policy-relevant intervention, home energy conservation reports. We measure consumer welfare by sending introductory reports and using an incentive-compatible multiple price list to determine willingness-to-pay to continue the program. We combine this with estimates of implementation costs and externality reductions to carry out a comprehensive welfare evaluation. We find that this nudge increases social welfare, although traditional program evaluation approaches overstate welfare gains by a factor of six. To exploit significant individual-level heterogeneity in welfare gains, we develop a simple machine learning algorithm to optimally target the nudge; this would increase welfare gains by more than 150 percent. Our results highlight that nudges, even those that are highly effective at changing behavior, need to be evaluated based on their welfare implications.