This paper provides a broad range of estimates of substitution elasticities for sectoral nested CES production functions, using panel data techniques, with the World Input-Output Database (WIOD) as the main data source. Although the related empirical literature has been growing over the recent years, there is still no single study focused on a large-scale estimation of various sectoral elasticities with a use of a common database and methodology. This paper constitutes an attempt to fill this gap. The obtained estimates may be subsequently used by Computable General Equilibrium (CGE) modellers in their applied research. A significant heterogeneity in estimated elasticity values is observed between various industries/products, as well as between various nests of the production function. This constitutes a strong argument against the arbitrary use of Leontief and/or Cobb- Douglas specifications in CGE models. It also turns out that, in most cases, obtained long-run elasticities are higher than in the short run. In addition, the analytical specification of estimated equation and time series properties of panel data (stationarity and cointegration) play a crucial role in determining the "correct" type of dynamic model (autoregressive distributed lag model, error correction model or model for differenced series) for a particular sector-nest combination and, in turn, in determining the preferred values of elasticity estimates.