Modeling the spatial aspect of growth has finally become an important subject of economics as exemplified by the increasing popularity of the new economic geography. However, new economic geography models have still not been able to develop a consistent approach to integrate innovation, space and economic growth into a coherent theoretical framework A potential reason for this is that the spatial dimension of knowledge production is still only partly understood in the empirical literature. To shed some additional light on the spatial dimension of innovation we present results of a first-cut analysis building on a recently developed cross sectional-time series data set of US innovation, private and university research and high technology employment. The novelty of this data set is that it opens up the possibilities to incorporate the time dimension into knowledge production function analysis at an appropriate level of spatial aggregation (i.e., US metropolitan areas) that has not been possible in empirical research yet.