The aim of this paper is to give an answer to the question, whether local spatial knowledge spillovers can be found in a cross section analysis for German NUTS-2 data. The analysis is based on regional production functions embedded in a general spatial model context. In addition, the paper expands the analysis to a Bayesian econometric view to allow for the existence of spatial heterogeneity in the data. Further, both using Bayesian and Non Bayesian methods, it should be more likely to obtain a more reliable model selection mechanism. Finally, using spatial filtering methods, own and neigbouring effects of regions regarding their innovative potential are separated. Particularly, the paper find evidence, first, that the output per capita of a region follows a spatial process, driven mainly by patents and human capital and second, that, consulting the knowledge production function theory, spatial knowledge spillovers clusters mainly exists with some exceptions in West German regions. Third, employing eigenvector based filter methods it was found that West German high productive regions are productive mainly due to their own innovative potential and mainly East German regions are confronted with negative neighbouring effects.