The purpose of this chapter is to discuss long-run elasticities of demand for gasoline in OECD countries. In so doing a number of alternative modelling strategies will be compared. The results can be understood as a further illustration of some of the methodological results on the inherent differences between various estimators analysed by Pesaran and Smith in Chapter 2. There are several reasons for choosing to model gasoline demand:

• Economically the most important petroleum products are the transport fuels. On the international scene these are the subject of two distinct but interconnected debates. On the one hand, there is the inevitable conflict between producers and consumers over prices. On the other hand there is the fact that increased transportation of goods and passengers causes various environmental problems, ranging from local congestion, noise and smog to global warming due to carbon (and other) emissions. Numerous technical, institutional and legal mechanisms are currently being assessed for their ability to mitigate such problems. For the economist the most natural approach is, presumably, to use the price mechanism in one form or another—be it through carbon taxes, road or zone fees, other special charges or simply gasoline taxes. On the issue of gasoline taxation and pricing in various countries, see Angelier and Sterner (1990) and Sterner (1989a, b). In the context of these debates, policy-makers in both producing and consuming countries have an obvious interest in obtaining good estimates of the elasticity of gasoline demand.

• Gasoline is also a relatively homogeneous product—at least compared with aggregate energy, which contains such inherently different energy carriers as oil, coal, wood and electricity. Homogeneity in this case makes it easier to compare between countries, to aggregate, to calculate average prices, and so forth.

• As a result of these two factors, gasoline demand has already been heavily researched, with a vast literature on this single subject. In a recent survey over a hundred published studies with several hundred individual estimates were discussed (see Dahl and Sterner 1991a, b). This survey showed that, although individual estimates do vary considerably, with a suitable stratification by type of data and methodology used the estimates become more similar. Since there are so many estimates and these estimates have such an immediate policy relevance there is also quite an interest in trying to resolve and explain the differences found. Therefore, in this study a large number of different modelling strategies have been applied on the same data.



The data used in this paper cover the OECD from 1960 to 1988 and come from the Gothenburg Energy Data Bank. Gasoline consumption is total (apparent) consumption of gasoline per capita, taken from UN data. Income is GDP per capita and prices are domestic consumer prices of regular gasoline, taken from the International Energy Agency and the US Energy Information Administration. All prices and incomes have been converted to a common base by using purchasing power parities from Summer and Heston (1988). And finally, vehicle data are from the International Road Federation.

Modelling strategies

A modeller must make a series of choices as to how to build his model. Some, but not all, of these choices can be guided by formal testing and not all of them can be discussed within the scope of a single chapter. Among the major decisions the modeller of gasoline demand faces are the selection of variables to be included, of appropriate functional form, and the level of aggregation. Among the many 'minor' decisions are, for instance, how to convert prices and incomes to a common currency.

This chapter discusses models that are already at a very aggregate level. The data set used is a panel of yearly data for a number of countries and years. It is therefore a large T panel or 'field' as defined in Chapter 2. The focus of the chapter is to compare time-series, cross-section and pooled estimators for this field. All the models discussed use the log-linear functional form and wherever necessary purchasing power parity conversions to constant US dollars are made.1 Furthermore, only two exogenous variables, gasoline prices and income levels, will be used.

The principal difficulty in modelling gasoline demand stems from the fact that it is a derived demand, depending heavily on the characteristics of the stock of relevant capital equipment, vehicles and transport infrastructure. But it also depends crucially on everyday decisions as to utilization. Thus, both current prices and incomes, as well as historic ones, have an influence on demand and the lags in adjustment to changing prices and incomes are long and may be complicated. One of the reasons for this is the omission of such variables as vehicle stock and infrastructure,2 which, in turn, have been affected by historic income levels and gasoline prices.

At the same time, however, it is found that there is relatively little variation in gasoline prices between years within each country (see for instance Sterner 1990). Most of the variation (for income, gasoline price and consumption alike) in the data set comes from the variation between countries. Some countries, such as the United States, have had 'cheap' gasoline throughout most of the time period studied, while others, such as Italy, have had 'expensive' gasoline. It thus seems reasonable, a priori, to expect that cross-sectional evidence will be needed to pick up effectively the long-run effects of adaptation to these two different price regimes.

Was this article helpful?

0 0

Post a comment