ANALYSES OF ENERGY DEMAND have invariably concentrated upon the estimation of key relationships, expressed as elasticities. There are two approaches to the calculation of income elasticities. The first, which we shall refer to as the engineering approach, measures income elasticity as the ratio of energy growth to gross domestic product growth; the energy demand studies canied out under this approach would place the emphasis on the dependence of energy demand on income. The second, which we shall refer to as the economic approach, would define the income elasticity on the hypothetical (ceteris paribus) marginal increase in energy demand, if the income (and income only) would change by, say, one per cent.The degree of sophistication applied to such studies has increased markedly over recent years. Thus, where earlier papers may have concentrated upon simple log-linear relationships, implying constant elasticities over the entire sample period, later attempts allow estimated parameters to move over time; see, for example, Kouris (1981), Bopp (1984), Abodunde et a1 (1985). Regardless of tests that are made concerning movements in elasticity over time, there is usually an implicit or explicit assumption that there exists a unique elasticity applicable to each explanatory variable at each mint in time. The classic test for changing parameters over time has concerned the nature of the link between income and energy demand, which appears to have changed following the 1970s' oil price adjustments. Elasticity is a dynamic marginal concept, in contrast to the energy demand/GDP ratio which is a static average concept. The argument developed below suggests that the level of growth of explanatory variables in that period can itself affect income elasticities. This paper concentrates on tests made for all OECD countries, to check whether such a relationship can be confirmed by the available data. Finally, an attempt is made to establish the implications of the results for other studies and forecasts.