Further evidence on social comparison and residential water use

In this paper, we evaluate a randomized controlled trial in which households were given Home Water Reports (HWRs) that provided recent consumption information aside a social comparison. We estimate an average treatment effect of −2.4%, consistent with previous literature. The effects are significantly larger during months that require irrigation but are still statistically significant during non-irrigation months. We then investigate if the treatment effect depends on the specific message a household receives. Using a regression discontinuity design, we find no evidence that the type of message received (e.g., “good” or “take action”) has an impact on the average response size of 2.4% in our particular setting. We use these findings to motivate a discussion regarding study design in the context of RCTs being used to identify heterogeneous impacts when they are not specifically designed to identify such effects. Specific to our study, we use ex post power tests to demonstrate that our findings related to message type are likely due to sample size.