by Fiona Burlig, James Bushnell, and David Rapson
When treatment effects are heterogeneous and experimental compliance is low, intent-to-treat (ITT) effects typically best approximate the population average treat- ment effect. In such circumstances, the most policy-relevant experiment will randomize all eligible subjects. In this paper, we randomize the population of electric vehicle (EV) owners in an EV-dense region of California into an EV managed charging program that controls the timing of EV load on the electric distribution system. Enrollment is below 5%, even among households that are offered the highest incentive of $40 per month. Contrary to local average treatment effects (LATEs) estimated in recent studies, our ITT effects on load are indistinguishable from zero. Results offer insights to policy- makers currently exploring strategies to integrate high levels of EVs into the electric grid, and reinforce the importance of randomizing among the eligible population.