The paper considers a neo-classical model set in the cost function approach to estimate primary energy factor demands for the Italian economy, using a translog cost function specification. Cointegration theory is employed to estimate the long-run factor share model, and the general to specific methodology to derive an error correction formulation for the short-run adjustment process. Both quarterly and yearly series, for the period 1978q1-1994q4 and 1960-1994, respectively, have been considered in the analysis. The different energy sources substitution pattern obtained by the quarterly and annual series and the super exogeneity property of the annual model suggest the importance of using low frequency data rather than quarterly data in estimating long-run relationships.