Non-linear technological progress and the substitutability of energy for capital: an application using the translog cost function.

This paper analyses the production process of four industries over four separate time periods using datasets taken form Berndt and Wood (1975, 1979), Hunt (1984a, 1986), Norsworthy and Harper (1981) and Jorgensen and Stiroh (2000). In their initial paper Berndt and Wood failed to explore the alternative options available to them to represent technological progress, a deficiency noted by Hunt (1986) who tested for alternative representations of technology (inter alia) using the Berndt and Wood data. This paper extends this line of reasoning/research by allowing technological progress to take more flexible non-linear forms using both deterministic and stochastic trend models. The results reveal that ‘non-linear trend’ models are generally preferred to ‘linear trend’ or ‘no trend’ models hence raising a question over the validity of assumptions used in much previous empirical research. Further the results reveal that the different assumptions lead to different results for the energy-capital elasticity of substitution.