Energy policy is increasingly focused on programs meant to counteract perceived underinvestment in energy efficiency. In evaluating these policies, economists have typically assumed that the actions of one household do not influence the choices of others. This misses a potentially important set of peer effects. If takeup of energy-effciency programs is constrained by information problems, social networks may be particularly important in determining participation. In addition, correctly assessing the cost effectiveness of these programs requires credible estimates of spillover benefits. Taking advantage of rich household-level data on one million appliance replacements in a large middle-income country, I use a regression discontinuity design to compare homes whose neighbors were barely eligible for an appliance replacement subsidy to homes whose neighbors were barely ineligible. I find evidence of substantial peer effects. The effect is concentrated within 1-2 months from the time of participation. The timing and size of the peer effect suggest that information problems may affect energy-effciency program participation.