A cointegration analysis of gasoline demand in the United States

Abstract

Time-series estimation of gasoline demand elasticities often does not take into account the possibility of nonstationarity in the underlying data, which may render the parameter estimates spurious. Studies have shown that the time trending variables used to explain gasoline demand could be difference stationary and therefore, may require cointegration analysis to assess the relationship among the trending variables. In this work we use the cointegration technique to derive long-run and short-run demand elasticities of noncommercial gasoline consumption using time-series data for the USA from 1949 to 2004. We also attempt to incorporate the presence of a structural break in the data generation process of the time trending variables. Our results show that the consumption of gasoline and lifetime income have a long-term stable relationship after the second oil shock of 1978. Prior to the first oil shock of 1973, no such long-run relationship could be established through cointegration.

Notes

¹Dahl and Sterner (1991), Sterner and Dahl (1992), Dahl (1995), Goodwin et al. (2004), de Jong and Gunn (2001) and Graham and Glaister (2002, 2004) – provided a survey of existing literature on fuel demand elasticities.

²One such spurious regression constructed by Hendry (1980) could explain that cumulative rainfall is a better indicator of price inflation than money stock in the UK!

³Ethanol is used as a motor fuel in Brazil.

⁴For example, a simple unit root test on G with a single lag involves testing for ρ=1 in the equation, G_t =γ+ρG_{t− 1}+(δt)+ε_t .

⁵Income appears to have a large lag-structure post-1978. This is behaviourally quite interesting as it may imply less income mobility between wage groups post-1978 as opposed to pre-1974.

⁶Long-run elasticities are calculated from β=λ/(1−λ_G ).