The question of the degree to which industrial consumers respond to fluctuations in the relative prices of energy and non-energy inputs has received considerable attention in recent years. Until the first major oil price shock of 1973/74, attention had focused primarily on capital-labour substitutability, and the effect this has had in the determination of income shares and economic growth. In the past, energy was not seen as a potential constraint to future growth and was generally excluded from analyses of factor substitution and distribution. Along with raw materials, it was regarded as a separable input which could be ignored in growth forecasting and macroeconomic projections.' Not surprisingly this has now changed. The role of depletable resources in present and future production and consumption possibilities of firms and individuals is seen as increasingly relevant. In particular, the question of whether consumption eventually approaches zero as resources are depleted, or whether growth is always feasible, has been analysed by several writers under assumptions concerning, for example, production techniques, differential extraction costs and the probability of developing backstop technologies.2 Fundamental to this is the issue of capital-resource and capital-energy substitutability. Not surprisingly, the feasibility of continued growth under energy constraints depends largely on whether capital and energy are substitutes. However the relevant empirical evidence is inconclusive. The relationship between energy and capital has been analysed in both econometric and engineering contexts. Berndt and Wood (1975) and Berndt and Jorgenson (1973) for U. S. industry, Fuss (1977) for Canada, Magnus (1979) for the Netherlands, and Swain and Friede (1976) for West Germany, all report energy-capital complementarity. Yet, Griffin and Gregory (1976) and Pindyck (1979), for the OECD area, and Cowing (1974), Ohta (1975) and Halvorsen and Ford (1979), for the U.S., report energycapital substitutability. Naturally these studies differ in terms of model * Manuscript received 22.11.84; final version received 2.6.87. tThe author would like to thank Robert Bacon, David Pear-, Michael Webb and an anonymous referee for comments. Any errors or sins of omission are his own responsibility.'An exception to this is the study by Barnett and Morse (1969) which emphasizes the role of natural mources and the importance of substitutability in the production process. %ee, for example, Dasgupta and Heal (1974) and Solow and Wan (1976).