In the current discussion on policy measures to alleviate competitive disadvantages of unilateral climate policies, border measures are discussed as one possible instrument. The assessment of the economic impacts of border measures is often carried out with the help of multi‐sector, multi‐region computable general equilibrium (CGE) models and depends crucially on empirical data and their quality, in particular if policy makers are to be supported in their concrete decisions. However, these models frequently rely on global economic databases which are, with respect to their sectoral coverage, in many cases too broad to account for specific climate change policies. For instance, border measures may have adverse impacts on energy‐intensive and trade‐exposed sectors subject to the risk of carbon leakage. The GTAP 7.1 database to which the CGE model PACE (Policy Analysis based on Computable Equilibrium) is calibrated treats many of these industries as part of larger aggregate sectors and might thus miss important information on the heterogeneity of these sectors. In this paper, we use the PACE model to investigate the potential merits of the disaggregation of selected economic sectors. We elaborate on the availability of data resources and methodological issues and make use of a harmonized dataset of supply and use tables with a high sectoral resolution to split the sub‐sectors "Cement, lime and plaster", "Aluminium products" and "Manufacturing of iron and steel" out of their respective aggregate GTAP sectors. Drawing on the example of border tax measures, we analyze the impacts of disaggregation on sub‐sectoral and macroeconomic indicators. Against the background of potential unobserved heterogeneity at the sub‐sectoral level, our main objective is to detect how sensitive CGE simulation results are in terms of changes in the parameterization of disaggregated GTAP sectors. Therefore, we perform sensitivity analyses that involve variations in trade elasticities, energy intensities and technology assumptions. First, we find that a sectoral classification which is too aggregate neglects important insights about sub‐sectoral implications. This shows the merits of sectoral disaggregation. Second, regarding the sensitivity analysis, we observe that tremendous deviations in the impact of border measures emerge from variations in Armington elasticities and the structure of the production functions. Third, the effects of sectoral disaggregation are not as pronounced for macroeconomic indicators and leakage as for sub‐sectoral indicators. Hence, if we are merely interested in macroeconomic impacts, the use of databases at a higher sectoral aggregation level is sufficient.