Text vs. citations: A comparative analysis of breakthrough and disruption metrics in patent innovation

Abstract

This study examines two dynamic metrics for assessing technological innovation— the text-based breakthrough index (KI index) and the citation-based disruption index (CD index)—both of which integrate ex-ante (novelty) and ex-post (impact) information. The KI index identifies breakthrough inventions by measuring their novelty (low similarity to prior patents) and impact (high similarity to future patents), whereas the CD index quantifies technological disruption by analyzing shifts in citation patterns. Using a dataset of over six million patents filed with the USPTO between 1980 and 2017, this paper finds that KI and CD indices are highly correlated and both effectively capture technological breakthroughs. Patents with high KI or CD scores typically originate from original and narrowly focused knowledge bases. However, the two indices exhibit distinct patterns: (1) the KI index fluctuates with economic cycles, while the CD index has experienced a steady decline over time; (2) the KI index positively correlates with future patent citation impact, whereas the CD index follows a U-shaped relationship with patent citation impact; and (3) small and remote teams produce higher KI but lower CD scores, potentially because larger teams cite newer, widely recognized references. I discuss innovation concepts—breakthroughs, disruptions, and beyond—to contextualize these findings and explore their implications for understanding technological advancement. These results contribute to the discourse on measuring innovation and underscore the complementary strengths of text-based and citation-based approaches in assessing technological progress.