Research collaboration and regional knowledge production in Europe

The focus of this study is on regional knowledge production activities in Europe, with special emphasis on the interplay between agglomeration and network effects. As increasingly considered in economic geography and regional science in the recent past, regional knowledge production activities, on the one hand, still remain geographically bounded; on the other hand, knowledge production activities have become increasingly interwoven and internationalized, emphasizing the crucial importance of region-external knowledge sources for a region's knowledge production capacity. The objective of the study is to estimate to what extent agglomeration and network effects influence knowledge production activities at the level of European regions. We use an extended regional knowledge production function framework as basis for the study, and derive a spatial Durbin model (SDM) relationship that can be used for empirical testing. The European coverage is achieved using 241 NUTS-2 regions covering the EU-25 member states. The dependent variable, knowledge production activity, is measured in terms of patent counts at the regional level in the time period 1998-2008, using patents applied at the European Patent Office (EPO). The independent variables include an agglomeration index, measured in terms of population density, and the regional participation intensity in the European network of R&D cooperation, measured in terms of the number of participations of a region in R&D joint ventures funded by the European Commission under the heading of the EU Framework programs (FPs). By this we are able to estimate the distinct effects of network participation and agglomeration on regional knowledge production. In our modeling framework, we further control for total regional R&D expenditures as widely used in regional knowledge production function frameworks and its empirical applications. In estimating the effects, we implement a panel version of the standard SDM that controls for spatial autocorrelation as well as individual heterogeneity across regions. The specification incorporates a spatial lag of the dependent variable as well as spatial lags of the independent variables. This allows for the estimation of spatial spillovers of agglomeration and network effects from neighboring regions by calculating scalar summary measures of impacts. The estimation results are expected to provide sketches of policy implications in a European and regional policy context.