The fact that innovative activity is remarkably concentrated in space, and in particular in cities, has motivated an important research effort to understand the spatial dimension of innovation, and the underlying mechanisms at work. While the literature has established the importance knowledge flows for location of innovation, the mechanisms through which they diffuse in space remain largely understudied. In particular, studies have insisted on the importance of skilled workers' mobility and the networked nature of knowledge production for innovation. Building on these considerations, this thesis investigates the role of skilled mobility in the diffusion of knowledge, and the resulting distribution of innovative activity. To answer this question, the thesis proceeds in three steps. The first chapter sets the conceptual framework and surveys the related literature. One of the main conclusion of this review is that some new economic geography and growth models provide a useful theoretical framework, because they recognize the importance of skilled mobility and knowledge externalities for the distribution of innovation. However, they fail to provide a reasonable answer to our research question for at least two reasons. First, the migration dynamics are very simplistic, and introducing heterogeneity in workers' characteristics and location preferences alters the cumulative mechanism of agglomeration. The second chapter provides a descriptive analysis on the patterns of inventors' mobility across urban areas, and their spatial dimension. Using these results, a spatial filtering gravity model is used to analyse formally how employment opportunities, professional networks and urban amenities, influence inventors' mobility flows. Second, these models do not consider workers' role in the diffusion of knowledge. The literature has established that skilled individuals influence the diffusion of knowledge by moving across organisation, creating network relationships and building absorptive capacities. The third chapter implements a spatial Durbin model to study these three mechanisms in an integrated framework. It is assumed that that mobility and networks provide access to knowledge, but the proportion of accessible knowledge used for innovation depends on absorptive capacity. These results have implications for the geography of innovation. While long-term mobility acts as a strong agglomeration force, the development of short-term, circular patterns of mobility should give rise to dispersion. The relative importance of these two effects is uncertain, because workers have different propensities and motivation to move, so that mobility patterns differ considerably. This should help explaining the persistence of long-run growth differentials among urban areas, and in a more dynamic setting, whether these gaps tend to widen or fall over time.