A randomized experiment is used to test whether employees will take actions to lower short- and long-run electricity use when their actions are unobservable and only the firm can benefit. Results suggest that social norms act as a coordinating device supporting worker conservation efforts. Electricity use fell 5.2% on average in buildings that were provided information on their own energy use compared to that in a paired building. The energy reductions have persisted over three years. Feedback on own past usage and provision of promotional information induced smaller and
statistically insignificant reductions in electricity use.