In a recent paper in this Review, Ernst Berndt and David Wood provide a useful clarification of and complementarity, pointing out that energy and capital can be substitutes in a production subfunction and yet complements in the aggregate production function. Using this theoretical proposition, they attempt to reconcile the econometric findings of energy-capital complementarity with other studies finding energy-capital substitutability. Their reconciliation rests on the finding that those studies finding energy-capital substitutability considered only capital (K), labor (L), and energy (E). By omitting materials (M) one obtains only a elasticity. The elasticity, which allows for the additional substitution between the KLE aggregate and M, can indicate energy-capital complementarity. In fact, KLEM studies generally find energy and capital complementarity. The purpose of this comment is to question whether the net and the gross elasticity distinction provides such a reconciliation among the econometric results. I agree that this explanation tends to reduce the disparity between the original Berndt-Wood elasticity estimate and Griffin-Gregory. Yet, three independent sources of evidence suggest that the omission of M is not a sufficient explanation, and probably not even a major explanation for the disparity of findings. First, the difference between the elasticity in a KLE submodel and the elasticity in a KLEM model depends critically on the elasticity of substitution between M and KLE. Let us adopt the Berndt-Wood notation and consider their mathematical example which demonstrates the possibility of capital energy substitutability as in the Griffin-Gregory study with capital-energy complementarity:

Results of earlier transcendental logarithmic (translog) production function are challenged on the basis of the limited time-series used. The same methodology is applied to a pool of international manufacturing data to test whether long-run price elasticities can be generated with intercountry samples. Both the time-series and international cross-section methods agree that there is some elasticity of substitution between energy demand and non-energy inputs and that energy forecasting should not be based on the assumption that substitutions between energy and nonenergy inputs are trivial. The international method finds, however, that capital and energy are substitutes rather than complements in the long-run and that there is a difference between short-run and long-run effects of substitution. 29 references. (DCK)