Abstract: Through exploratory spatial data analysis, the spatial agglomeration characteristics of outputs and inputs of Chinese provincial knowledge production are found. Considering spatial spillover effects of outputs and inputs, a modified knowledge production spatial panel data model is established. By means of LM test and Wald test, it is found that spatial Durbin model is the best model for describing the activities of provincial knowledge production. The empirical research discoveries that knowledge outputs have significant positive spillover effects; R&D active people have significant negative spillover effects; R&D expenditure and FDI have positive spillover effects, however, are not very significant. Based on the results, the suggestions for improving the efficiency of provincial knowledge production are put forwards.

Key words: knowledge spillover, knowledge production, spatial panel data model, spatial Durbin model