Richard Schmalensee and Thomas M. Stoker, Econometrica, Vol. 67, No. 3, pp. 645-62 (May 1999)
Continued rapid growth in U.S. gasoline consumption is of particular interest because of various environmental consequences, from increased urban pollution and congestion to overall climate change. For projection trends in gasoline consumption, one can turn to an extensive econometric literature on gasoline demand, based largely on aggregate data. These studies tend to find a long-run income elasticity of around unity, which suggests substantial future growth in gasoline consumption in the U.S. and abroad.
In this paper, we give results from studying household-level data on gasoline consumption. Our analysis is motivated by several issues that arise from using existing studies to project secular trends in gasoline consumption. First is the question of whether high income households display the same income elasticity as other households. Will households with incomes of $60,000 really drive about twice as much on average as households with incomes of $30,000? If not, the aggregate income elasticity of demand may well fall over time, with consumption growth slowing relative to GDP growth. The most natural way to learn about the gasoline demand of future high-income consumers is to study the behavior of today\'s high-income consumers, which leads us to using household-level data.
Second, over periods of a decade or more, age structures and other demographic characteristics may change substantially, and it is reasonable to expect such changes will affect gasoline demand. The United States has seen a substantial aging of the population over the last several decades, as well as the emergence of two-earner (often two-commuter) households as a typical model. However, as Dahl (1993) notes, very little work has been done on gasoline demand at the household level, and even less has taken full account of differences in household composition and location. Moreover, secular trends are associated with longer-run determinants of demand, which are also consistent with the use of household-level data.
