Lawrence Berkeley National Laboratory

During the period 1973 to 1985, the U.S. economy saved energy in virtually every sector. Much of this period of energy saving was also marked by a significant drop in the ratio of energy use to GDP. However, since 1985 there has been a slowdown in the rate of energy saving, as key energy intensities (space heating, automobile driving, etc.) have declined less rapidly since 1985 than before. This paper examines delivered (or final) energy consumption trends from the early 1970s to 1994 and provides a framework for measuring key changes that affect U.S. energy use. Starting with estimates of outputs or activity levels for thirty major energy end uses, and energy intensities of each end use, we use the Adaptive Weighted Divisia decomposition to measure the impact of changes in the structure of the U.S. economy. In contrast to many similar decomposition studies, we define measures of structural changes for both households and branches of transportation. We find that between 1973 and 1985, lower energy intensities (corrected to average winter heating demand) reduced U.S. energy uses by about 1.7 percent per year, while structural changes reduced energy uses by 0.4 percent per year. After 1985, when oil prices declined markedly, intensities fell by only 0.8 percent per year and structural changes actually increased energy use by 0.4 percent per year. In the 1990s energy intensities in some industries have even edged upward. Changes in the ratio of energy to GDP (E/GDP) are affected both by intensities and the changes in the demand for energy services relative to GDP. During the first period, from 1973 to 1985, GDP increased faster than the growth in key structural and activity parameters that determine demand for energy services (such as home area, appliance ownership, and motor vehicle use) by 1.5 percent per year. From 1985 to 1994 the difference dropped to less than 0.3 percent per year, largely due to the reversal of structural trends. Thus, the sharp fall in the rate of decline in E/GDP from -3.1 percent to -1.1 percent per year was due almost as much to structural changes as it was to the slowdown in energy intensity reduction. The analysis presented here shows why the E/GDP is an increasingly unreliable yardstick for making measurements of how the energy-economy relationship is changing: effects not related to energy efficiency per se may have roughly the same impact on that ratio as energy saving itself. Since these effects have different causes, and potentially different impacts over the long run, looking at them in the aggregate by considering only the ratio of energy use to GDP is misleading.