Energy intensity of the economy is often modeled as being determined by the combined effect of a fixed price elasticity of demand, and an exogenously specified, fixed technical change parameter denoted as the autonomous energy efficiency improvement (AEEI). Typically, the AEEI rate is set to 0.5-1.5% improvement per annum. Here, we study historic aggregate energy intensity trends for the US from 1954 to 1994. We show that the historic trends are inconsistent with an autonomous model of improved energy efficiency-especially when the model is used to inform policies that impact energy prices. As an alternative we propose a model of price-induced efficiency, pi, in which aggregate energy intensity trends respond to changes in energy prices beyond price elasticity of demand epsilon. Our exercise reveals that the aggregate price elasticity of energy demand of the US economy has declined by roughly 15% over the past four decades. But beyond this decline, bringing our simulations and historical data into close correspondence requires pi to change sign before and after 1974. Before 1974, after accounting for price elasticity of demand, the economy was growing less energy efficient. After 1974, after accounting for the price elasticity of demand, the economy was growing more energy efficient. Furthermore, since 1984, the rate of energy efficiency gain has been declining. When projections of long-term energy use are compared, those with a price-induced energy efficiency formulation generate significantly more price sensitive energy use and emissions trajectories. When in the business as usual scenario energy prices are expected to be rising, climate policies involve lower shadow carbon prices with pi than with AEEI formulations. In scenarios where energy prices are relatively flat, energy intensity rises leading to CO2 emissions far higher than standard business as usual projections utilizing AEEI assumptions. (c) 2005 Elsevier Ltd. All rights reserved.