Abstract
One of the central questions in the debate about agglomeration economies is the effect of the density of economic activity on productivity and wages. To advance the understanding of this process, we employ a very rich plant-level dataset covering all economic activity in Sweden with building-level geographic precision. The density of economic activity is computed on various spatial scales using nonparametric methods. Its elasticity in relation to wages is estimated at 1.7%, considerably lower than most studies using spatially aggregated data. Further, spatial variables are constructed for individual agglomeration mechanisms, including market and supplier access, thickness of local labour markets and a proxy for knowledge spillovers. The individual mechanisms all have strong effects even after controlling for general density, though their relative importance varies depending on industry characteristics.