What trends in energy efficiencies? Evidence from a robust test

A proper modeling of the long-run behavior of energy and oil intensities is crucial in many respects. This paper aims at checking whether this long-run behavior should be modelled as a deterministic or a stochastic trend or both. We first apply a test for a deterministic trend robust to uncertainty about the stochastic trend. Our results indicate that, for the period 1960-2004, energy intensities of only 8 OECD countries out of 25 include a negative deterministic trend, 3 include a positive one and 14 seem to be better modelled by a stochastic trend only. When considering a sample of 73 non-OECD countries on the period 1971-2004, we show that only 22 exhibit a deterministic trend (negative for 15 countries and positive for 7 countries). A similar analysis for oil intensity leads to reject the hypothesis of an insignificant deterministic trend for 7 OECD countries out of 23 for the period 1965-2004 and 11 non-OECD countries out of 40 for the period 1971-2004. In the next step, we apply standard unit root tests and find that the unit root hypothesis is not very often rejected. We conclude that a main feature of energy intensities is the presence of a stochastic trend.