Knowledge Networks from Patent Data

Abstract

The economic literature on technical change has increasingly relied upon patent citation data to measure inter-personal knowledge flows. Many doubts exist about whether patent citations really reflect the designated inventors’ knowledge of both their technical fields, and of the other inventors and experts therein: citations, in fact, come mainly from the patent examiners, and possibly the patent applicant’s lawyers, rather than from inventors themselves. Unfortunately, most of the papers dedicated to discussing these interpretation issues deal with USPTO data, whose citation rules are quite exceptional if compared to those of other patent offices. In addition some confusion exists between the two issues of awareness (whether citing inventors actually knew of the cited patents) and existence of a knowledge flow (whether some information on the contents of the cited patents has however reached the, possibly unaware, citing inventor). Questionnaires addressed to inventors are severely affected by this confusion, and can hardly dispel the existing doubts. We then propose to apply social network analysis to derive maps of social relationships between inventors, and measures of social proximity between cited and citing patents. Logit regressions demonstrate that the probability of observing a citation is positively influenced by such proximity. In order to perform such regressions, however, a specific sampling scheme has to used, which we also illustrate and discuss.