This paper specifies and estimates state-level models of short- and long-term electricity demand in the United States. The short-term model predicts hourly load based on weather and calendar inputs. The long-term model estimates interannual demand, and includes population, prices, and gross state product as predictors. These models are combined to incorporate the short- and long-term trends in electricity consumption when generating forecasts of diurnal patterns into the future. Finally, the authors investigate the effects of short-run price elasticities of demand. The short-term model is shown to be within 95% accuracy of actual levels in out-of-sample tests. (C) 2012 Elsevier Ltd. All rights reserved.