There is a considerable disagreement in the literature on the magnitude of primary energy use and CO2 emissions linked to the production of buildings. In this paper we assess the Swedish building sector using top-down input-output analysis. These top-down results are then disaggregated into sectors and activities, which are compared to results from 18 previous bottom-up studies using process-LCA methodology. The analysis shows almost 90% higher specific energy use (GJ/m(2)) for the top-down methodology. The differences are only around 20% for the share coupled to production and processing of building materials, while for other involved sectors such as transport, construction activities, production of machines and service sectors, the input output analysis gives much higher values. Some of these differences can be explained by truncation errors due to the definition of system boundaries in the bottom-up studies. The apparent underestimation of energy use for transport, services etc. in bottom-up studies is only of marginal importance when comparing for example materials choices, but when comparing the production phase to the use phase of buildings such errors are likely to result in an underestimation of the relative importance of the production phase since the use phase is dominated by more easily estimated direct energy use. (c) 2007 Elsevier Ltd. All rights reserved.