Beyond the front page: In‐text citations to patents as traces of inventor knowledge

Abstract

Abstract Research Summary This study introduces in‐text patent‐to‐patent citations—references embedded in the body of patent documents—as a novel data source to trace knowledge flows. Unlike front‐page citations, which often reflect legal requirements, in‐text citations are more likely to originate from inventors and signal meaningful technological linkages. We show that they exhibit stronger geographic and semantic proximity, greater self‐referentiality, and closer alignment with inventor knowledge. Though less frequent than front‐page citations, they yield robust results in models of knowledge diffusion. We release a validated dataset and reproducible code to support future research. Our findings offer new opportunities for scholars interested in the microfoundations of innovation, the geography of knowledge flows, and the role of inventors in shaping firms' knowledge trajectories. Managerial Summary This study introduces in‐text patent citations—references embedded in the technical description of patent documents—as a new way to trace how knowledge flows among inventors and firms. Unlike traditional front‐page citations, which often reflect legal formalities or examiner input, in‐text citations originate more directly from inventors and thus capture genuine technological linkages. Using large‐scale U.S. patent data, we show that in‐text citations connect geographically and thematically closer inventions and are more likely to involve the same firms or inventors. The dataset, released as open data, enables more accurate analyses of innovation dynamics and firm knowledge strategies. Managers can use these insights to better understand how ideas diffuse and where valuable inventive knowledge resides.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 6 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.