Demand-side Management (DSM) programs by electricity utilities report substantial energy savings that often receive little support from empirical studies. We argue that this discrepancy results from an inherently static view of technology adoption by utilities when estimating future energy savings. We illustrate this through a simple model of technology adoption, in which households operate different vintages of appliances and have heterogenous forecasts about the rate of future technological progress. An "energy efficiency gap" arises when households under-estimate the true rate of technological progress. We parameterize the model using data on refrigerators and show that a DSM program that subsidizes adoption of energy-efficient refrigerators yields small energy saving that, in most cases, do not justify the cost of the subsidy.