The network structure of success: Evidence from an empirical study of European patents.

Patents are often used as a source of data to study innovation, and one measure of the "success" of a patent is the number of citations it receives from other patents, which is its in-degree in the patent citation network. The combination of different knowledge is the basis for innovation, which, almost by definition, entails the combination of knowledge in novel ways. But not all possible combinations of knowledge are equally likely to succeed. What factors contribute to the success of a patent? In this work we use the ideas of categorical contrast and niche width to help try to answer this question. The contrast of a category (such as a patent technology class) captures the idea of sharp versus broad or "fuzzy" category boundaries. If a technology class is one that is seldom assigned together with other classes, then it has high contrast. A low contrast technology class is one that is frequently found together with other classes. Niche width is a measure of the diversity of technology classes combined by a patent. We will use these ideas to examine the effects of diversity and contrast on the success of a patent, using a data set of nearly two million European patents. We will use both conventional regression modeling, as well as exponential random graph modeling (ERGM) to model the patent citation network. The latter allows modeling of the patent citation network without having to treat the network structure as exogenous.This work was funded by Swiss National Science Foundation NRP 75 Big Data project 167326 "The Global Structure of Knowledge Networks: Data, Models and Empirical Results