In-text patent citations: A user's guide

• Published in: Research policy Vol. 49; no. 4; p. 103946
• Main Authors: Bryan, Kevin A., Ozcan, Yasin, Sampat, Bhaven
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.05.2020
• ISSN: 0048-7333; 1873-7625
• DOI: 10.1016/j.respol.2020.103946

•Introduces a new dataset of in-text patent citations to science.•Shows that there is little overlap between in-text and front-page citations.•The journals and type of science cited in-text is different than on the front page.•In-text citations are much more likely to appear in initial patent application.•In-text citations more closely track other measures of inventor knowledge. We introduce, validate, and make publicly available a new data source for innovation research: scientific references in patent specifications. These references are common and algorithmically extractable. Critically, they are very different from the “front page” prior art commonly used to proxy for inventor knowledge. Only 24% of front page citations to academic articles are in the patent text, and 31% of in-text citations are on the front page. We explain these differences by describing the legal rules and practice governing citation. Empirical validations suggest that in-text citations are qualitatively different. Further, there is suggestive evidence that they more accurately proxy for knowledge flows, consistent with their legal role.