Many countries use substantial public funds to subsidize reductions in negative externalities. Such policy designs create asymmetric incentives because increases in externalities remain unpriced. I investigate the implications of such policies by using a regression discontinuity design in California's electricity rebate program. Using household-level panel data, I find that the incentive produced precisely estimated zero treatment effects on energy conservation in coastal areas. In contrast, the rebate induced short-run and long-run consumption reductions in inland areas. Income, climate, and air conditioner saturation significantly drive the heterogeneity. Finally, I provide a cost-effectiveness analysis and investigate how to improve the policy design. (JEL D12, D62, H76, L94, L98, Q48)