Distributional and Efficiency Impacts of Gasoline Taxes: An Econometrically Based Multi-market Study

摘要

Because of its potential to improve the environment and enhance national security, reducing automobile-related gasoline consumption has become a major U.S. public policy issue. Recently, many analysts have called for new or more stringent policies to discourage gasoline consumption. Proposals include a tightening of corporate average fuel economy (CAFE) standards and subsidies to retirements of older (gas-guzzling) vehicles, as well as increments to the federal gasoline tax. This paper examines the gas-tax option, employing an econometrically based multi-market simulation model to explore the policy's efficiency and distributional implications. This study differs from earlier work in several ways. Some prior studies have investigated gasoline consumption either by employing a demand function for gasoline or by deriving this demand from households' vehicle-miles traveled (VMT). These studies treat the composi- tion of the automobile fleet as fixed. However, a gasoline tax can be expected to influence the fleet composition (e.g., the market share of more fuel-efficient cars) as well as the amount of driving. This study allows for both impacts. As in Steven Berry et al. (1995), Pinelopi Goldberg (1995), and Amil Petrin (2002), we account for the imperfectly competitive nature of the new-car market However, in contrast with these studies, we consider interactions between the markets for new, used, and scrapped cars. The impacts of a gasoline tax can importantly depend on such interactions. Higher gasoline taxes could stimulate higher rates of scrappage of older, fuel-inefficient cars and could also promote shifts in demand from used cars to especially fuel-efficient new cars. Studies that ignore these adjustments could understate a gas tax's impacts on fuel consumption. Another set of differences from earlier work is in the econometric approach. By allowing the structural parameters entering preferences to vary randomly across households, we can account for rich patterns of unobserved preference heterogeneity. Also, in contrast with nearly all prior work, we adopt an estimation approach that simultaneously estimates in a utility-consistent manner each household's automobile choice and its choice of VMT. This is important for evaluating welfare impacts.