Heterogeneous treatment effects and mechanisms in information-based environmental policies: Evidence from a large-scale field experiment

Policymakers often rely on non-pecuniary, information-based programs to achieve social objectives. Using data from a water conservation information campaign implemented as a randomized controlled trial, we estimate heterogeneous household responses. Understanding such heterogeneity is important for improving the cost-effectiveness of non-pecuniary programs, extending them to other populations and probing the mechanisms through which the treatment effects arise. We find little evidence of heterogeneous responses to purely technical information or to traditional conservation messages that combine technical information and moral suasion. In contrast, norm-based messages that combine technical information, moral suasion and social comparisons exhibit strong heterogeneity: households that are wealthier, owner-occupied and use more water are more responsive. These subgroups tend to be least responsive to pecuniary incentives. We find no evidence that any subgroup increases their water use in response to the messages. By targeting the messages to subgroups known to be most responsive, program costs could be reduced by over 50% with only a 20% reduction in the treatment effect. Combining theory and data, we also shed light on the mechanisms through which the treatment effects arise, which has implications for program design and future research on the program's welfare effects.