* The author would like to thank Russell Sobel, Yoram Barzel, George Hammond, and participants at the 2003 Southern Economic Association meeting for helpful comments. All remaining errors are the responsibility of the author. Excise Taxation and Product Quality: The Gasoline Market* Todd M. Nesbit Department of Economics College of Business & Economics P.O. Box 6025 West Virginia University Morgantown, WV 26506 (304) 293-7861 [email protected] Abstract Following Barzel (1976), product quality increases in response to unit taxation but remains unchanged by ad valorem taxation. While many tax theorists agree this argument is theoretically sound, empirical support of Barzel’s theory is limited to the cigarette market. This paper tests and confirms his theory in the gasoline market, a market in which Barzel failed to find supporting evidence in his original article. Using a direct test proposed by Sobel and Garrett (1997) and improved data, I find the market shares of premium and mid-grade gasoline rise in response to per-unit taxation but are unaffected by ad valorem taxation.1 Excise Taxation and Product Quality: The Gasoline Market I. Introduction Standard tax theory shows that as the tax rate rises, individuals will purchase less of the taxed good or resource. This quantity substitution, however, is not the only margin on which individuals can act to minimize their tax liability. Barzel (1976) explains that individuals may also alter their purchases of quality in response to changes in the tax rate. He theorized that for unit taxation individuals will substitute away from the taxed attribute of the good (which is quantity) and into its untaxed attribute (which is quality). Barzel’s theory suggests that while both per-unit and ad valorem taxes lead to a decline in product sales, per-unit taxes will also lead to an increase in average product quality. An alternative explanation of this quality substitution can be drawn from the Alchian and Allen (1967) theorem because the imposition of a per-unit tax (a fixed fee) causes the price of the higher-quality version of a good to fall relative to a lower-quality substitute. Ad valorem taxes, however, do not alter the relative prices of the different quality levels of the product and, thus, do not impact the purchase of quality.1 While Barzel’s idea is both novel and widely cited, it has received limited support empirically, with nearly all of the tests having been conducted in the cigarette market. Indeed, only Barzel’s original article attempted to test his theory in a market other than cigarettes.2 While he found empirical support for his theory in the cigarette market, his results were inconclusive in the markets for gasoline and alcohol. The goal of this paper is to reexamine the gasoline market using data that was not available when Barzel first analyzed this market, allowing for a better empirical model. This reexamination of2 Barzel’s theory is important because there is a need to determine whether the theory is unique to the cigarette market or whether it can be applied more generally. As will be explained in the next section of the paper, Barzel’s model is inherently biased toward rejecting his theory, which may explain his inability to find evidence of quality substitution in the market for gasoline. The method used in this paper corrects this bias. The paper proceeds as follows. In the next section, the previous empirical literature is reviewed, making note of the drawbacks of Barzel’s original method and the benefits of the model used here. Section III presents the empirical model used in this study and discusses the results. Concluding remarks are offered in section IV. II. A Review of the Empirical Literature Barzel originally tested his hypothesis in the cigarette market using data from 1954 to 1972. While his hypothesis was that state per unit cigarette taxes should increase the quality of cigarettes, this was impossible to test directly without objective data on the quality of cigarettes. Instead, Barzel relied on an indirect test using the price of cigarettes. His model assumed that the full burden of the tax would fall upon consumers, causing the retail price of cigarettes to increase by the full amount of the tax. Thus, if the imposition of a unit tax also causes an increase in product quality, the price change resulting from the unit tax should be greater than the amount of the unit tax itself (reflecting also the value of the quality improvement). This can be illustrated by the following equation: ∆P = T + ∆V3 where ∆P is the observed after-tax price change, T is the unit tax, and ∆V is the value of the change in product quality. The following numerical example may aid in this discussion. If a twenty cent unit tax caused the price to rise by twenty-five cents, Barzel would attribute this difference to an increase in product quality valued at five cents. However, Barzel’s results are critically sensitive to his assumption that the full burden of the unit tax falls on consumers. If consumers bear less than the full burden of the tax, which depends on the relative elasticities of demand and supply, increases in quality may go unmeasured in Barzel’s model. Consider the same twenty cent unit tax and same increase in product quality valued at five cents discussed above, but assume instead that consumers and producers equally share the burden of the tax. The observed price increase is only fifteen cents, potentially leading to the rejection of the theory when it is actually true. Barzel tested his theory with the following empirical model: where Pit is the average retail price per pack of cigarettes in state i and year t, TAXit is the per unit tax on cigarettes in state i and year t, NHTAXit is the product of a dummy variable for New Hampshire with TAXit, DISTANCEi is the road distance from North Carolina to the largest metropolitan area in state i, and also included in the regression is a set of year dummy variables. Despite the potential bias of the model used by Barzel, data limitations forced him to adopt such an indirect test of his theory. Barzel’s test of his hypothesis that unit taxation leads to increased product quality centered on whether the coefficient on the TAX variable is greater than one. εααααiti3it2it10it + dummies year + DISTANCE + NHTAX + TAX + = P4 Additionally, to test the hypothesis that