Dynamic Formation of Knowledge Networks and Innovating Firm

Abstract

We document a series of new facts about the very first firms and patents that form new edges in the directed citation networks across patent categories. We call them pathfinder firms and patents. First, the typical pathfinder firms are very larger firms. Second, the average pathfinder patents have higher quality than other patents in different quality measures. Third, firms innovate faster, market value, profit and productivity increase in the future, when they invent a large number of pathfinder patents or are cited by pathfinder patents currently. Fourth, new citation links generate positive externalities to peer firms that innovate in nearby technology space. We then build a dynamic formation model of knowledge network, where new citations form through both quality based preferential attachment, exact and mutated copying of parent patent’s citations. The model rationalizes the firm level empirical facts. After calibrated to the patent citation data, the model sheds lights on the causes of productivity show down: rising number of citation per patent, marginal cost of patent quality and research wage have driven up the cost to innovate since 1976. Effective R&D policies need to decrease these cost factors.