We estimate agglomeration economies, defined as the effect of density on labour productivity in European regions. The analysis of Ciccone (2002) is extended in two main ways. First, we use dynamic panel estimation techniques (system GMM), thus offering an alternative methodological treatment of the inherent endogeneity problem. Second, the sector dimension in the data allows for disaggregated estimation. Our results confirm the presence of significant agglomeration effects at the aggregate level, with an estimated long-run elasticity of 13 percent. Repeated crosssection regressions suggest that the strength of agglomeration effects has increased over time. At the sector level, the dominant pattern is of cross-sector "urbanisation" economies and own-sector congestion diseconomies. A notable exception is financial services, for which we find strong positive productivity effects from own-sector density.