This paper investigates the effect of disaggregate energy consumption on industrial output in the United States. Most of the related research utilizes aggregate data which may not indicate the relative strength or explanatory power of various energy inputs on output. We use monthly data and employ the generalized variance decomposition approach to assess the relative impacts of energy and employment on real output. Our results suggest that unexpected shocks to coal, natural gas and fossil fuel energy sources have the highest impacts on the variation of output, while several renewable sources exhibit considerable explanatory power as well. However, none of the energy sources explain more of the forecast error variance of industrial output than employment. (c) 2006 Elsevier Ltd. All rights reserved.